1 | <?xml version="1.0" encoding="utf-8" standalone="no"?>
|
---|
2 | <!DOCTYPE Archive SYSTEM "http://greenstone.org/dtd/Archive/1.0/Archive.dtd">
|
---|
3 | <Archive>
|
---|
4 | <Section>
|
---|
5 | <Description>
|
---|
6 | <Metadata name="gsdldoctype">indexed_doc</Metadata>
|
---|
7 | <Metadata name="Language">en</Metadata>
|
---|
8 | <Metadata name="Encoding">windows_1252</Metadata>
|
---|
9 | <Metadata name="Creator">dg5</Metadata>
|
---|
10 | <Metadata name="Title">1997-00 Listing of Working Papers</Metadata>
|
---|
11 | <Metadata name="URL">http://C:/Users/Anupama/GS307_13July2015/web/sites/localsite/collect/Word-PDF-Enhanced/tmp/1436775751/word01.html</Metadata>
|
---|
12 | <Metadata name="UTF8URL">http://C:/Users/Anupama/GS307_13July2015/web/sites/localsite/collect/Word-PDF-Enhanced/tmp/1436775751/word01.html</Metadata>
|
---|
13 | <Metadata name="gsdlsourcefilename">import\word01.doc</Metadata>
|
---|
14 | <Metadata name="gsdlconvertedfilename">tmp\1436775751\word01.html</Metadata>
|
---|
15 | <Metadata name="OrigSource">word01.html</Metadata>
|
---|
16 | <Metadata name="Source">word01.doc</Metadata>
|
---|
17 | <Metadata name="SourceFile">word01.doc</Metadata>
|
---|
18 | <Metadata name="Plugin">WordPlugin</Metadata>
|
---|
19 | <Metadata name="FileSize">110080</Metadata>
|
---|
20 | <Metadata name="FilenameRoot">word01</Metadata>
|
---|
21 | <Metadata name="FileFormat">Word</Metadata>
|
---|
22 | <Metadata name="srcicon">_icondoc_</Metadata>
|
---|
23 | <Metadata name="srclink_file">doc.doc</Metadata>
|
---|
24 | <Metadata name="srclinkFile">doc.doc</Metadata>
|
---|
25 | <Metadata name="Identifier">HASHeaa2992e081949673150f3</Metadata>
|
---|
26 | <Metadata name="lastmodified">1436763858</Metadata>
|
---|
27 | <Metadata name="lastmodifieddate">20150713</Metadata>
|
---|
28 | <Metadata name="oailastmodified">1436775752</Metadata>
|
---|
29 | <Metadata name="oailastmodifieddate">20150713</Metadata>
|
---|
30 | <Metadata name="assocfilepath">HASHeaa2.dir</Metadata>
|
---|
31 | <Metadata name="gsdlassocfile">doc.doc:application/msword:</Metadata>
|
---|
32 | </Description>
|
---|
33 | <Content>
|
---|
34 |
|
---|
35 |
|
---|
36 |
|
---|
37 | <div class=WordSection1>
|
---|
38 |
|
---|
39 |
|
---|
40 |
|
---|
41 | <p class=MsoTitle><span lang=EN-US>1997-00 Listing of Working Papers </span></p>
|
---|
42 |
|
---|
43 |
|
---|
44 |
|
---|
45 |
|
---|
46 |
|
---|
47 |
|
---|
48 |
|
---|
49 |
|
---|
50 |
|
---|
51 |
|
---|
52 |
|
---|
53 |
|
---|
54 |
|
---|
55 |
|
---|
56 |
|
---|
57 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/1</span></p>
|
---|
58 |
|
---|
59 |
|
---|
60 |
|
---|
61 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Using
|
---|
62 |
|
---|
63 | compression to identify acronyms in text</span></p>
|
---|
64 |
|
---|
65 |
|
---|
66 |
|
---|
67 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Stuart <span
|
---|
68 |
|
---|
69 | class=SpellE>Yeates</span>, David Bainbridge, Ian H. <span class=SpellE>Witten</span></span></p>
|
---|
70 |
|
---|
71 |
|
---|
72 |
|
---|
73 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Text mining is
|
---|
74 |
|
---|
75 | about looking for patterns in natural language text, and may be defined as the
|
---|
76 |
|
---|
77 | process of <span class=SpellE>analyzing</span> text to extract information from
|
---|
78 |
|
---|
79 | it for particular purposes.<span style='mso-spacerun:yes'>Â </span>In previous
|
---|
80 |
|
---|
81 | work, we claimed that compression is a key technology for text mining, and
|
---|
82 |
|
---|
83 | backed this up with a study that showed how particular kinds of lexical
|
---|
84 |
|
---|
85 | tokensânames, dates, locations, <i style='mso-bidi-font-style:normal'>etc.</i>âcan
|
---|
86 |
|
---|
87 | be identified and located in running text, using compression models to provide
|
---|
88 |
|
---|
89 | the leverage necessary to distinguish different token types (Witten <i
|
---|
90 |
|
---|
91 | style='mso-bidi-font-style:normal'>et al.</i>, 1999)</span></p>
|
---|
92 |
|
---|
93 |
|
---|
94 |
|
---|
95 |
|
---|
96 |
|
---|
97 |
|
---|
98 |
|
---|
99 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/2</span></p>
|
---|
100 |
|
---|
101 |
|
---|
102 |
|
---|
103 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Text <span
|
---|
104 |
|
---|
105 | class=SpellE>categorization</span> using compression models</span></p>
|
---|
106 |
|
---|
107 |
|
---|
108 |
|
---|
109 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
110 |
|
---|
111 | lang=EN-GB>Eibe</span></span><span lang=EN-GB> Frank, Chang <span class=SpellE>Chui</span>,
|
---|
112 |
|
---|
113 | Ian H. <span class=SpellE>Witten</span></span></p>
|
---|
114 |
|
---|
115 |
|
---|
116 |
|
---|
117 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Text <span
|
---|
118 |
|
---|
119 | class=SpellE>categorization</span>, or the assignment of natural language texts
|
---|
120 |
|
---|
121 | to predefined categories based on their content, is of growing importance as
|
---|
122 |
|
---|
123 | the volume of information available on the internet continues to overwhelm
|
---|
124 |
|
---|
125 | us.<span style='mso-spacerun:yes'>Â </span>The use of predefined categories implies
|
---|
126 |
|
---|
127 | a âsupervised learningâ approach to <span class=SpellE>categorization</span>,
|
---|
128 |
|
---|
129 | where already-classified articles â which effectively define the categories â
|
---|
130 |
|
---|
131 | are used as âtraining dataâ to build a model that can be used for classifying
|
---|
132 |
|
---|
133 | new articles that comprise the âtest dataâ.<span style='mso-spacerun:yes'>Â
|
---|
134 |
|
---|
135 | </span>This contrasts with âunsupervisedâ learning, where there is no training
|
---|
136 |
|
---|
137 | data and clusters of like documents are sought amongst the test articles.<span
|
---|
138 |
|
---|
139 | style='mso-spacerun:yes'>Â </span>With supervised learning, meaningful labels
|
---|
140 |
|
---|
141 | (such as <span class=SpellE>keyphrases</span>) are attached to the training
|
---|
142 |
|
---|
143 | documents, and appropriate labels can be assigned automatically to test
|
---|
144 |
|
---|
145 | documents depending on which category they fall into.</span></p>
|
---|
146 |
|
---|
147 |
|
---|
148 |
|
---|
149 |
|
---|
150 |
|
---|
151 |
|
---|
152 |
|
---|
153 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/3</span></p>
|
---|
154 |
|
---|
155 |
|
---|
156 |
|
---|
157 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Reserved for
|
---|
158 |
|
---|
159 | Sally Jo</span></p>
|
---|
160 |
|
---|
161 |
|
---|
162 |
|
---|
163 |
|
---|
164 |
|
---|
165 |
|
---|
166 |
|
---|
167 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/4</span></p>
|
---|
168 |
|
---|
169 |
|
---|
170 |
|
---|
171 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Interactive
|
---|
172 |
|
---|
173 | machine learningâletting users build classifiers</span></p>
|
---|
174 |
|
---|
175 |
|
---|
176 |
|
---|
177 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Malcolm Ware, <span
|
---|
178 |
|
---|
179 | class=SpellE>Eibe</span> Frank, Geoffrey Holmes, Mark Hall, Ian H. <span
|
---|
180 |
|
---|
181 | class=SpellE>Witten</span></span></p>
|
---|
182 |
|
---|
183 |
|
---|
184 |
|
---|
185 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>According to
|
---|
186 |
|
---|
187 | standard procedure, building a classifier is a fully automated process that
|
---|
188 |
|
---|
189 | follows data preparation by a domain expert.<span style='mso-spacerun:yes'>Â
|
---|
190 |
|
---|
191 | </span>In contrast, &lt;I&gt;interactive&lt;/I&gt;machine learning engages
|
---|
192 |
|
---|
193 | users in actually generating the classifier themselves.<span
|
---|
194 |
|
---|
195 | style='mso-spacerun:yes'>Â </span>This offers a natural way of integrating
|
---|
196 |
|
---|
197 | background knowledge into the <span class=SpellE>modeling</span> stageâso long
|
---|
198 |
|
---|
199 | as interactive tools can be designed that support efficient and effective
|
---|
200 |
|
---|
201 | communication.<span style='mso-spacerun:yes'>Â </span>This paper shows that
|
---|
202 |
|
---|
203 | appropriate techniques can empower users to create models that compete with
|
---|
204 |
|
---|
205 | classifiers built by state-of-the-art learning algorithms.<span
|
---|
206 |
|
---|
207 | style='mso-spacerun:yes'>Â </span>It demonstrates that usersâeven users who are
|
---|
208 |
|
---|
209 | not domain expertsâcan often construct good classifiers, without any help from
|
---|
210 |
|
---|
211 | a learning algorithm, using a simple two-dimensional visual interface.<span
|
---|
212 |
|
---|
213 | style='mso-spacerun:yes'>Â </span>Experiments demonstrate that, not
|
---|
214 |
|
---|
215 | surprisingly, success hinges on the domain: if a few attributes can support
|
---|
216 |
|
---|
217 | good predictions, users generate accurate classifiers, whereas domains with
|
---|
218 |
|
---|
219 | many high-order attribute interactions <span class=SpellE>favor</span> standard
|
---|
220 |
|
---|
221 | machine learning techniques.<span style='mso-spacerun:yes'>Â </span>The future
|
---|
222 |
|
---|
223 | challenge is to achieve a symbiosis between human user and machine learning
|
---|
224 |
|
---|
225 | algorithm.</span></p>
|
---|
226 |
|
---|
227 |
|
---|
228 |
|
---|
229 |
|
---|
230 |
|
---|
231 |
|
---|
232 |
|
---|
233 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/5</span></p>
|
---|
234 |
|
---|
235 |
|
---|
236 |
|
---|
237 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>KEA: Practical
|
---|
238 |
|
---|
239 | automatic <span class=SpellE>keyphrase</span> extraction</span></p>
|
---|
240 |
|
---|
241 |
|
---|
242 |
|
---|
243 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Ian H. <span
|
---|
244 |
|
---|
245 | class=SpellE>Witten</span>, Gordon W. <span class=SpellE>Paynter</span>, <span
|
---|
246 |
|
---|
247 | class=SpellE>Eibe</span> Frank, Carl <span class=SpellE>Gutwin</span>, Craig G.
|
---|
248 |
|
---|
249 | <span class=SpellE>Nevill</span>-Manning</span></p>
|
---|
250 |
|
---|
251 |
|
---|
252 |
|
---|
253 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
254 |
|
---|
255 | lang=EN-GB>Keyphrases</span></span><span lang=EN-GB> provide semantic metadata
|
---|
256 |
|
---|
257 | that <span class=SpellE>summarize</span> and <span class=SpellE>characterize</span>
|
---|
258 |
|
---|
259 | documents.<span style='mso-spacerun:yes'>Â </span>This paper describes <span
|
---|
260 |
|
---|
261 | class=SpellE>Kea</span>, an algorithm for automatically extracting <span
|
---|
262 |
|
---|
263 | class=SpellE>keyphrases</span> from text.<span style='mso-spacerun:yes'>Â
|
---|
264 |
|
---|
265 | </span><span class=SpellE>Kea</span> identifies candidate <span class=SpellE>keyphrases</span>
|
---|
266 |
|
---|
267 | using lexical methods, calculates feature values for each candidate, and uses a
|
---|
268 |
|
---|
269 | machine learning algorithm to predict which candidates are good <span
|
---|
270 |
|
---|
271 | class=SpellE>keyphrases</span>.<span style='mso-spacerun:yes'>Â </span>The
|
---|
272 |
|
---|
273 | machine learning scheme first builds a prediction model using training
|
---|
274 |
|
---|
275 | documents with known <span class=SpellE>keyphrases</span>, and then uses the
|
---|
276 |
|
---|
277 | model to find <span class=SpellE>keyphrases</span> in new documents.<span
|
---|
278 |
|
---|
279 | style='mso-spacerun:yes'>Â </span>We use a large test corpus to evaluate <span
|
---|
280 |
|
---|
281 | class=SpellE>Kea's</span> effectiveness in terms of how many author-assigned <span
|
---|
282 |
|
---|
283 | class=SpellE>keyphrases</span> are correctly identified.<span
|
---|
284 |
|
---|
285 | style='mso-spacerun:yes'>Â </span>The system is simple, robust, and publicly
|
---|
286 |
|
---|
287 | available.</span></p>
|
---|
288 |
|
---|
289 |
|
---|
290 |
|
---|
291 |
|
---|
292 |
|
---|
293 |
|
---|
294 |
|
---|
295 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/6</span></p>
|
---|
296 |
|
---|
297 |
|
---|
298 |
|
---|
299 | <p class=MsoNormal style='margin-right:-.4pt'><i style='mso-bidi-font-style:
|
---|
300 |
|
---|
301 | normal'><span lang=EN-GB style='font-family:Symbol;mso-ascii-font-family:"Times New Roman";
|
---|
302 |
|
---|
303 | mso-hansi-font-family:"Times New Roman";mso-char-type:symbol;mso-symbol-font-family:
|
---|
304 |
|
---|
305 | Symbol'><span style='mso-char-type:symbol;mso-symbol-font-family:Symbol'>m</span></span></i><span
|
---|
306 |
|
---|
307 | lang=EN-GB>-Charts and Z:<span style='mso-spacerun:yes'>Â </span><span
|
---|
308 |
|
---|
309 | class=SpellE>hows</span>, <span class=SpellE>whys</span> and <span
|
---|
310 |
|
---|
311 | class=SpellE>wherefores</span></span></p>
|
---|
312 |
|
---|
313 |
|
---|
314 |
|
---|
315 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Greg Reeve,
|
---|
316 |
|
---|
317 | Steve Reeves</span></p>
|
---|
318 |
|
---|
319 |
|
---|
320 |
|
---|
321 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>In this paper we
|
---|
322 |
|
---|
323 | show, by a series of examples, how the </span><i style='mso-bidi-font-style:
|
---|
324 |
|
---|
325 | normal'><span lang=EN-GB style='font-family:Symbol;mso-ascii-font-family:"Times New Roman";
|
---|
326 |
|
---|
327 | mso-hansi-font-family:"Times New Roman";mso-char-type:symbol;mso-symbol-font-family:
|
---|
328 |
|
---|
329 | Symbol'><span style='mso-char-type:symbol;mso-symbol-font-family:Symbol'>m</span></span></i><span
|
---|
330 |
|
---|
331 | lang=EN-GB>-chart formalism can be translated into Z.<span
|
---|
332 |
|
---|
333 | style='mso-spacerun:yes'>Â </span>We give reasons for why this is an
|
---|
334 |
|
---|
335 | interesting and sensible thing to do and what it might be used for.</span></p>
|
---|
336 |
|
---|
337 |
|
---|
338 |
|
---|
339 |
|
---|
340 |
|
---|
341 |
|
---|
342 |
|
---|
343 |
|
---|
344 |
|
---|
345 |
|
---|
346 |
|
---|
347 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/7</span></p>
|
---|
348 |
|
---|
349 |
|
---|
350 |
|
---|
351 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>One dimensional
|
---|
352 |
|
---|
353 | non-uniform rational B-splines for animation control</span></p>
|
---|
354 |
|
---|
355 |
|
---|
356 |
|
---|
357 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
358 |
|
---|
359 | lang=EN-GB>Abdelaziz</span></span><span lang=EN-GB> <span class=SpellE>Mahoui</span></span></p>
|
---|
360 |
|
---|
361 |
|
---|
362 |
|
---|
363 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Most 3D
|
---|
364 |
|
---|
365 | animation packages use graphical representations called motion graphs to
|
---|
366 |
|
---|
367 | represent the variation in time of the motion parameters.<span
|
---|
368 |
|
---|
369 | style='mso-spacerun:yes'>Â </span>Many use two-dimensional B-splines as
|
---|
370 |
|
---|
371 | animation curves because of their power to represent free-form curves.<span
|
---|
372 |
|
---|
373 | style='mso-spacerun:yes'>Â </span>In this project, we investigate the
|
---|
374 |
|
---|
375 | possibility of using One-dimensional Non-Uniform Rational B-<span class=SpellE>Spline</span>
|
---|
376 |
|
---|
377 | (NURBS) curves for the interactive construction of animation control
|
---|
378 |
|
---|
379 | curves.<span style='mso-spacerun:yes'>Â </span>One-dimensional NURBS curves
|
---|
380 |
|
---|
381 | present the potential of solving some problems encountered in motion graphs
|
---|
382 |
|
---|
383 | when two-dimensional B-splines are used.<span style='mso-spacerun:yes'>Â
|
---|
384 |
|
---|
385 | </span>The study focuses on the properties of One-dimensional NURBS
|
---|
386 |
|
---|
387 | mathematical model.<span style='mso-spacerun:yes'>Â </span>It also investigates
|
---|
388 |
|
---|
389 | the algorithms and shape modification tools devised for two-dimensional curves
|
---|
390 |
|
---|
391 | and their port to the One-dimensional NURBS model.<span
|
---|
392 |
|
---|
393 | style='mso-spacerun:yes'>Â </span>It also looks at the issues related to the
|
---|
394 |
|
---|
395 | user interface used to interactively modify the shape of the curves.</span></p>
|
---|
396 |
|
---|
397 |
|
---|
398 |
|
---|
399 |
|
---|
400 |
|
---|
401 |
|
---|
402 |
|
---|
403 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/8</span></p>
|
---|
404 |
|
---|
405 |
|
---|
406 |
|
---|
407 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Correlation-based
|
---|
408 |
|
---|
409 | feature selection of discrete and numeric class machine learning</span></p>
|
---|
410 |
|
---|
411 |
|
---|
412 |
|
---|
413 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Mark A. Hall</span></p>
|
---|
414 |
|
---|
415 |
|
---|
416 |
|
---|
417 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Algorithms for
|
---|
418 |
|
---|
419 | feature selection fall into two broad categories:
|
---|
420 |
|
---|
421 | &lt;I&gt;wrappers&lt;/I&gt;that use the learning algorithm itself to evaluate
|
---|
422 |
|
---|
423 | the usefulness of features and &lt;I&gt;filters&lt;/I&gt;that evaluate features
|
---|
424 |
|
---|
425 | according to heuristics based on general characteristics of the data.<span
|
---|
426 |
|
---|
427 | style='mso-spacerun:yes'>Â </span>For application to large databases, filters
|
---|
428 |
|
---|
429 | have proven to be more practical than wrappers because they are much
|
---|
430 |
|
---|
431 | faster.<span style='mso-spacerun:yes'>Â </span>However, most existing filter
|
---|
432 |
|
---|
433 | algorithms only work with discrete classification problems.<span
|
---|
434 |
|
---|
435 | style='mso-spacerun:yes'>Â </span>This paper describes a fast,
|
---|
436 |
|
---|
437 | correlation-based filter algorithm that can be applied to continuous and
|
---|
438 |
|
---|
439 | discrete problems.<span style='mso-spacerun:yes'>Â </span>The algorithm often
|
---|
440 |
|
---|
441 | out-performs the well-known <span class=SpellE>ReliefF</span> attribute
|
---|
442 |
|
---|
443 | estimator when used as a <span class=SpellE>preprocessing</span> step for naïve
|
---|
444 |
|
---|
445 | <span class=SpellE>Bayes</span>, instance-based learning, decision trees,
|
---|
446 |
|
---|
447 | locally weighted regression, and model trees.<span style='mso-spacerun:yes'>Â
|
---|
448 |
|
---|
449 | </span>It performs more feature selection than <span class=SpellE>ReliefF</span>
|
---|
450 |
|
---|
451 | does-reducing the data dimensionality by fifty percent in most cases.<span
|
---|
452 |
|
---|
453 | style='mso-spacerun:yes'>Â </span>Also, decision and model trees built from the
|
---|
454 |
|
---|
455 | <span class=SpellE>prepocessed</span> data are often significantly smaller.</span></p>
|
---|
456 |
|
---|
457 |
|
---|
458 |
|
---|
459 |
|
---|
460 |
|
---|
461 |
|
---|
462 |
|
---|
463 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/9</span></p>
|
---|
464 |
|
---|
465 |
|
---|
466 |
|
---|
467 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>A development
|
---|
468 |
|
---|
469 | environment for predictive modelling in foods</span></p>
|
---|
470 |
|
---|
471 |
|
---|
472 |
|
---|
473 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>G. Holmes, <span
|
---|
474 |
|
---|
475 | class=SpellE>M.A.</span> Hall</span></p>
|
---|
476 |
|
---|
477 |
|
---|
478 |
|
---|
479 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>WEKA (Waikato
|
---|
480 |
|
---|
481 | Environment for Knowledge Analysis) is a comprehensive suite of Java class
|
---|
482 |
|
---|
483 | libraries that implement many state-of-the-art machine learning/data mining
|
---|
484 |
|
---|
485 | algorithms.<span style='mso-spacerun:yes'>Â </span>Non-programmers interact
|
---|
486 |
|
---|
487 | with the software via a user interface component called the Knowledge Explorer.</span></p>
|
---|
488 |
|
---|
489 |
|
---|
490 |
|
---|
491 |
|
---|
492 |
|
---|
493 |
|
---|
494 |
|
---|
495 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Applications
|
---|
496 |
|
---|
497 | constructed from the WEKA class libraries can be run on any computer with a web
|
---|
498 |
|
---|
499 | browsing capability, allowing users to apply machine learning techniques to
|
---|
500 |
|
---|
501 | their own data regardless of computer platform.<span style='mso-spacerun:yes'>Â
|
---|
502 |
|
---|
503 | </span>This paper describes the user interface component of the WEKA system in
|
---|
504 |
|
---|
505 | reference to previous applications in the predictive <span class=SpellE>modeling</span>
|
---|
506 |
|
---|
507 | of foods.</span></p>
|
---|
508 |
|
---|
509 |
|
---|
510 |
|
---|
511 |
|
---|
512 |
|
---|
513 |
|
---|
514 |
|
---|
515 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/10</span></p>
|
---|
516 |
|
---|
517 |
|
---|
518 |
|
---|
519 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Benchmarking
|
---|
520 |
|
---|
521 | attribute selection techniques for data mining</span></p>
|
---|
522 |
|
---|
523 |
|
---|
524 |
|
---|
525 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Mark A. Hall,
|
---|
526 |
|
---|
527 | Geoffrey Holmes</span></p>
|
---|
528 |
|
---|
529 |
|
---|
530 |
|
---|
531 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Data engineering
|
---|
532 |
|
---|
533 | is generally considered to be a central issue in the development of data mining
|
---|
534 |
|
---|
535 | applications.<span style='mso-spacerun:yes'>Â </span>The success of many
|
---|
536 |
|
---|
537 | learning schemes, in their attempts to construct models of data, hinges on the
|
---|
538 |
|
---|
539 | reliable identification of a small set of highly predictive attributes.<span
|
---|
540 |
|
---|
541 | style='mso-spacerun:yes'>Â </span>The inclusion of irrelevant, redundant and
|
---|
542 |
|
---|
543 | noisy attributes in the model building process phase can result in poor
|
---|
544 |
|
---|
545 | predictive performance and increased computation.</span></p>
|
---|
546 |
|
---|
547 |
|
---|
548 |
|
---|
549 |
|
---|
550 |
|
---|
551 |
|
---|
552 |
|
---|
553 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Attribute
|
---|
554 |
|
---|
555 | selection generally involves a combination of search and attribute utility
|
---|
556 |
|
---|
557 | estimation plus evaluation with respect to specific learning schemes.<span
|
---|
558 |
|
---|
559 | style='mso-spacerun:yes'>Â </span>This leads to a large number of possible
|
---|
560 |
|
---|
561 | permutations and has led to a situation where very few benchmark studies have
|
---|
562 |
|
---|
563 | been conducted.</span></p>
|
---|
564 |
|
---|
565 |
|
---|
566 |
|
---|
567 |
|
---|
568 |
|
---|
569 |
|
---|
570 |
|
---|
571 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
572 |
|
---|
573 | presents a benchmark comparison of several attribute selection methods.<span
|
---|
574 |
|
---|
575 | style='mso-spacerun:yes'>Â </span>All the methods produce an attribute ranking,
|
---|
576 |
|
---|
577 | a useful devise of isolating the individual merit of an attribute.<span
|
---|
578 |
|
---|
579 | style='mso-spacerun:yes'>Â </span>Attribute selection is achieved by
|
---|
580 |
|
---|
581 | cross-validating the rankings with respect to a learning scheme to find the
|
---|
582 |
|
---|
583 | best attributes.<span style='mso-spacerun:yes'>Â </span>Results are reported
|
---|
584 |
|
---|
585 | for a selection of standard data sets and two learning schemes C4.5 and naïve <span
|
---|
586 |
|
---|
587 | class=SpellE>Bayes</span>.</span></p>
|
---|
588 |
|
---|
589 |
|
---|
590 |
|
---|
591 |
|
---|
592 |
|
---|
593 |
|
---|
594 |
|
---|
595 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/11</span></p>
|
---|
596 |
|
---|
597 |
|
---|
598 |
|
---|
599 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Steve Reeves,
|
---|
600 |
|
---|
601 | Greg Reeve</span></p>
|
---|
602 |
|
---|
603 |
|
---|
604 |
|
---|
605 |
|
---|
606 |
|
---|
607 |
|
---|
608 |
|
---|
609 |
|
---|
610 |
|
---|
611 |
|
---|
612 |
|
---|
613 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>2000/12</span></p>
|
---|
614 |
|
---|
615 |
|
---|
616 |
|
---|
617 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
618 |
|
---|
619 | lang=EN-GB>Malika</span></span><span lang=EN-GB> <span class=SpellE>Mahoui</span>,
|
---|
620 |
|
---|
621 | Sally Jo Cunningham</span></p>
|
---|
622 |
|
---|
623 |
|
---|
624 |
|
---|
625 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Transaction logs
|
---|
626 |
|
---|
627 | are invaluable sources of fine-grained information about users' search <span
|
---|
628 |
|
---|
629 | class=SpellE>behavior</span>.<span style='mso-spacerun:yes'>Â </span>This paper
|
---|
630 |
|
---|
631 | compares the searching <span class=SpellE>behavior</span> of users across two
|
---|
632 |
|
---|
633 | WWW-accessible digital libraries: the New Zealand Digital Library's Computer
|
---|
634 |
|
---|
635 | Science Technical Reports collection (CSTR), and the <span class=SpellE>Karlsruhe</span>
|
---|
636 |
|
---|
637 | Computer Science Bibliographies (CSBIB) collection.<span
|
---|
638 |
|
---|
639 | style='mso-spacerun:yes'>Â </span>Since the two collections are designed to
|
---|
640 |
|
---|
641 | support the same type of users-researchers/students in computer science a
|
---|
642 |
|
---|
643 | comparative log analysis is likely to uncover common searching preferences for
|
---|
644 |
|
---|
645 | that user group.<span style='mso-spacerun:yes'>Â </span>The two collections
|
---|
646 |
|
---|
647 | differ in their content, however; the CSTR indexes a full text collection,
|
---|
648 |
|
---|
649 | while the CSBIB is primarily a bibliographic database.<span
|
---|
650 |
|
---|
651 | style='mso-spacerun:yes'>Â </span>Differences in searching <span class=SpellE>behavior</span>
|
---|
652 |
|
---|
653 | between the two systems may indicate the effect of differing search facilities
|
---|
654 |
|
---|
655 | and content type.</span></p>
|
---|
656 |
|
---|
657 |
|
---|
658 |
|
---|
659 |
|
---|
660 |
|
---|
661 |
|
---|
662 |
|
---|
663 |
|
---|
664 |
|
---|
665 |
|
---|
666 |
|
---|
667 |
|
---|
668 |
|
---|
669 |
|
---|
670 |
|
---|
671 |
|
---|
672 |
|
---|
673 |
|
---|
674 |
|
---|
675 |
|
---|
676 |
|
---|
677 |
|
---|
678 |
|
---|
679 |
|
---|
680 |
|
---|
681 |
|
---|
682 |
|
---|
683 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/1</span></p>
|
---|
684 |
|
---|
685 |
|
---|
686 |
|
---|
687 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Lexical
|
---|
688 |
|
---|
689 | attraction for text compression</span></p>
|
---|
690 |
|
---|
691 |
|
---|
692 |
|
---|
693 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
694 |
|
---|
695 | lang=EN-GB>Joscha</span></span><span lang=EN-GB> Bach, Ian H. <span
|
---|
696 |
|
---|
697 | class=SpellE>Witten</span></span></p>
|
---|
698 |
|
---|
699 |
|
---|
700 |
|
---|
701 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>New methods of
|
---|
702 |
|
---|
703 | acquiring structural information in text documents may support better
|
---|
704 |
|
---|
705 | compression by identifying an appropriate prediction context for each
|
---|
706 |
|
---|
707 | symbol.<span style='mso-spacerun:yes'>Â </span>The method of âlexical
|
---|
708 |
|
---|
709 | attractionâ infers syntactic dependency structures from statistical analysis of
|
---|
710 |
|
---|
711 | large corpora.<span style='mso-spacerun:yes'>Â </span>We describe the
|
---|
712 |
|
---|
713 | generation of a lexical attraction model, discuss its application to text
|
---|
714 |
|
---|
715 | compression, and explore its potential to outperform fixed-context models such
|
---|
716 |
|
---|
717 | as word-level PPM.<span style='mso-spacerun:yes'>Â </span>Perhaps the most
|
---|
718 |
|
---|
719 | exciting aspect of this work is the prospect of using compression as a metric
|
---|
720 |
|
---|
721 | for structure discovery in text.</span></p>
|
---|
722 |
|
---|
723 |
|
---|
724 |
|
---|
725 |
|
---|
726 |
|
---|
727 |
|
---|
728 |
|
---|
729 |
|
---|
730 |
|
---|
731 |
|
---|
732 |
|
---|
733 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/2</span></p>
|
---|
734 |
|
---|
735 |
|
---|
736 |
|
---|
737 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Generating rule
|
---|
738 |
|
---|
739 | sets from model trees</span></p>
|
---|
740 |
|
---|
741 |
|
---|
742 |
|
---|
743 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Geoffrey Holmes,
|
---|
744 |
|
---|
745 | Mark Hall, <span class=SpellE>Eibe</span> Frank</span></p>
|
---|
746 |
|
---|
747 |
|
---|
748 |
|
---|
749 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Knowledge discovered
|
---|
750 |
|
---|
751 | in a database must be represented in a form that is easy to understand.<span
|
---|
752 |
|
---|
753 | style='mso-spacerun:yes'>Â </span>Small, easy to interpret nuggets of knowledge
|
---|
754 |
|
---|
755 | from data are one requirement and the ability to induce them from a variety of
|
---|
756 |
|
---|
757 | data sources is a second.<span style='mso-spacerun:yes'>Â </span>The literature
|
---|
758 |
|
---|
759 | is abound with classification algorithms, and in recent years with algorithms
|
---|
760 |
|
---|
761 | for time sequence analysis, but relatively little has been published on
|
---|
762 |
|
---|
763 | extracting meaningful information from problems involving continuous classes
|
---|
764 |
|
---|
765 | (regression).</span></p>
|
---|
766 |
|
---|
767 |
|
---|
768 |
|
---|
769 |
|
---|
770 |
|
---|
771 |
|
---|
772 |
|
---|
773 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Model
|
---|
774 |
|
---|
775 | trees-decision trees with linear models at the leaf nodes-have recently emerged
|
---|
776 |
|
---|
777 | as an accurate method for numeric prediction that produces understandable
|
---|
778 |
|
---|
779 | models.<span style='mso-spacerun:yes'>Â </span>However, it is well known that
|
---|
780 |
|
---|
781 | decision lists-ordered sets of If-Then rules-have the potential to be more compact
|
---|
782 |
|
---|
783 | and therefore more understandable than their tree counterparts.</span></p>
|
---|
784 |
|
---|
785 |
|
---|
786 |
|
---|
787 |
|
---|
788 |
|
---|
789 |
|
---|
790 |
|
---|
791 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>In this paper we
|
---|
792 |
|
---|
793 | present an algorithm for inducing simple, yet accurate rule sets from model
|
---|
794 |
|
---|
795 | trees.<span style='mso-spacerun:yes'>Â </span>The algorithm works by repeatedly
|
---|
796 |
|
---|
797 | building model trees and selecting the best rule at each iteration.<span
|
---|
798 |
|
---|
799 | style='mso-spacerun:yes'>Â </span>It produces rule sets that are, on the whole,
|
---|
800 |
|
---|
801 | as accurate but smaller than the model tree constructed from the entire <span
|
---|
802 |
|
---|
803 | class=SpellE>dataset</span>.<span style='mso-spacerun:yes'>Â
|
---|
804 |
|
---|
805 | </span>Experimental results for various heuristics which attempt to find a
|
---|
806 |
|
---|
807 | compromise between rule accuracy and rule coverage are reported.<span
|
---|
808 |
|
---|
809 | style='mso-spacerun:yes'>Â </span>We also show empirically that our method
|
---|
810 |
|
---|
811 | produces more accurate and smaller rule sets than the commercial
|
---|
812 |
|
---|
813 | state-of-the-art rule learning system Cubist.</span></p>
|
---|
814 |
|
---|
815 |
|
---|
816 |
|
---|
817 |
|
---|
818 |
|
---|
819 |
|
---|
820 |
|
---|
821 |
|
---|
822 |
|
---|
823 |
|
---|
824 |
|
---|
825 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/3</span></p>
|
---|
826 |
|
---|
827 |
|
---|
828 |
|
---|
829 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>A diagnostic
|
---|
830 |
|
---|
831 | tool for tree based supervised classification learning algorithms</span></p>
|
---|
832 |
|
---|
833 |
|
---|
834 |
|
---|
835 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Leonard <span
|
---|
836 |
|
---|
837 | class=SpellE>Trigg</span>, Geoffrey Holmes</span></p>
|
---|
838 |
|
---|
839 |
|
---|
840 |
|
---|
841 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The process of
|
---|
842 |
|
---|
843 | developing applications of machine learning and data mining that employ
|
---|
844 |
|
---|
845 | supervised classification algorithms includes the important step of knowledge
|
---|
846 |
|
---|
847 | verification.<span style='mso-spacerun:yes'>Â </span>Interpretable output is
|
---|
848 |
|
---|
849 | presented to a user so that they can verify that the knowledge contained in the
|
---|
850 |
|
---|
851 | output makes sense for the given application.<span style='mso-spacerun:yes'>Â
|
---|
852 |
|
---|
853 | </span>As the development of an application is an iterative process it is quite
|
---|
854 |
|
---|
855 | likely that a user would wish to compare models constructed at various times or
|
---|
856 |
|
---|
857 | stages.</span></p>
|
---|
858 |
|
---|
859 |
|
---|
860 |
|
---|
861 |
|
---|
862 |
|
---|
863 |
|
---|
864 |
|
---|
865 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>One crucial
|
---|
866 |
|
---|
867 | stage where comparison of models is important is when the accuracy of a model
|
---|
868 |
|
---|
869 | is being estimated, typically using some form of cross-validation.<span
|
---|
870 |
|
---|
871 | style='mso-spacerun:yes'>Â </span>This stage is used to establish an estimate
|
---|
872 |
|
---|
873 | of how well a model will perform on unseen data.<span
|
---|
874 |
|
---|
875 | style='mso-spacerun:yes'>Â </span>This is vital information to present to a
|
---|
876 |
|
---|
877 | user, but it is also important to show the degree of variation between models
|
---|
878 |
|
---|
879 | obtained from the entire <span class=SpellE>dataset</span> and models obtained
|
---|
880 |
|
---|
881 | during cross-validation.<span style='mso-spacerun:yes'>Â </span>In this way it
|
---|
882 |
|
---|
883 | can be verified that the cross-validation models are at least structurally
|
---|
884 |
|
---|
885 | aligned with the model garnered from the entire <span class=SpellE>dataset</span>.</span></p>
|
---|
886 |
|
---|
887 |
|
---|
888 |
|
---|
889 |
|
---|
890 |
|
---|
891 |
|
---|
892 |
|
---|
893 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
894 |
|
---|
895 | presents a diagnostic tool for the comparison of tree-based supervised
|
---|
896 |
|
---|
897 | classification models.<span style='mso-spacerun:yes'>Â </span>The method is
|
---|
898 |
|
---|
899 | adapted from work on approximate tree matching and applied to decision
|
---|
900 |
|
---|
901 | trees.<span style='mso-spacerun:yes'>Â </span>The tool is described together
|
---|
902 |
|
---|
903 | with experimental results on standard <span class=SpellE>datasets</span>.</span></p>
|
---|
904 |
|
---|
905 |
|
---|
906 |
|
---|
907 |
|
---|
908 |
|
---|
909 |
|
---|
910 |
|
---|
911 |
|
---|
912 |
|
---|
913 |
|
---|
914 |
|
---|
915 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/4</span></p>
|
---|
916 |
|
---|
917 |
|
---|
918 |
|
---|
919 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Feature
|
---|
920 |
|
---|
921 | selection for discrete and numeric class machine learning</span></p>
|
---|
922 |
|
---|
923 |
|
---|
924 |
|
---|
925 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Mark A. Hall</span></p>
|
---|
926 |
|
---|
927 |
|
---|
928 |
|
---|
929 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Algorithms for
|
---|
930 |
|
---|
931 | feature selection fall into two broad categories:
|
---|
932 |
|
---|
933 | &lt;I&gt;wrappers&lt;/I&gt;use the learning algorithm itself to evaluate the
|
---|
934 |
|
---|
935 | usefulness of features, while &lt;I&gt;filters&lt;/I&gt;evaluate features
|
---|
936 |
|
---|
937 | according to heuristics based on general characteristics of the data.<span
|
---|
938 |
|
---|
939 | style='mso-spacerun:yes'>Â </span>For application to large databases, filters
|
---|
940 |
|
---|
941 | have proven to be more practical than wrappers because they are much
|
---|
942 |
|
---|
943 | faster.<span style='mso-spacerun:yes'>Â </span>However, most existing filter
|
---|
944 |
|
---|
945 | algorithms only work with discrete classification problems.</span></p>
|
---|
946 |
|
---|
947 |
|
---|
948 |
|
---|
949 |
|
---|
950 |
|
---|
951 |
|
---|
952 |
|
---|
953 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
954 |
|
---|
955 | describes a fast, correlation-based filter algorithm that can be applied to
|
---|
956 |
|
---|
957 | continuous and discrete problems.<span style='mso-spacerun:yes'>Â
|
---|
958 |
|
---|
959 | </span>Experiments using the new method as a <span class=SpellE>preprocessing</span>
|
---|
960 |
|
---|
961 | step for naïve <span class=SpellE>Bayes</span>, instance-based learning,
|
---|
962 |
|
---|
963 | decision trees, locally weighted regression, and model trees show it to be an
|
---|
964 |
|
---|
965 | effective feature selector- it reduces the data in dimensionality by more than
|
---|
966 |
|
---|
967 | sixty percent in most cases without negatively affecting accuracy.<span
|
---|
968 |
|
---|
969 | style='mso-spacerun:yes'>Â </span>Also, decision and model trees built from the
|
---|
970 |
|
---|
971 | pre-processed data are often significantly smaller.</span></p>
|
---|
972 |
|
---|
973 |
|
---|
974 |
|
---|
975 |
|
---|
976 |
|
---|
977 |
|
---|
978 |
|
---|
979 |
|
---|
980 |
|
---|
981 |
|
---|
982 |
|
---|
983 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/5</span></p>
|
---|
984 |
|
---|
985 |
|
---|
986 |
|
---|
987 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Browsing tree
|
---|
988 |
|
---|
989 | structures</span></p>
|
---|
990 |
|
---|
991 |
|
---|
992 |
|
---|
993 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Mark <span
|
---|
994 |
|
---|
995 | class=SpellE>Apperley</span>, Robert <span class=SpellE>Spence</span>, Stephen <span
|
---|
996 |
|
---|
997 | class=SpellE>Hodge</span>, Michael Chester</span></p>
|
---|
998 |
|
---|
999 |
|
---|
1000 |
|
---|
1001 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Graphic
|
---|
1002 |
|
---|
1003 | representations of tree structures are notoriously difficult to create,
|
---|
1004 |
|
---|
1005 | display, and interpret, particularly when the volume of information they
|
---|
1006 |
|
---|
1007 | contain, and hence the number of nodes, is large.<span
|
---|
1008 |
|
---|
1009 | style='mso-spacerun:yes'>Â </span>The problem of interactively browsing
|
---|
1010 |
|
---|
1011 | information held in tree structures is examined, and the implementation of an
|
---|
1012 |
|
---|
1013 | innovative tree browser described.<span style='mso-spacerun:yes'>Â </span>This
|
---|
1014 |
|
---|
1015 | browser is based on distortion-oriented display techniques and intuitive direct
|
---|
1016 |
|
---|
1017 | manipulation interaction.<span style='mso-spacerun:yes'>Â </span>The tree
|
---|
1018 |
|
---|
1019 | layout is automatically generated, but the location and extent of detail shown
|
---|
1020 |
|
---|
1021 | is controlled by the user.<span style='mso-spacerun:yes'>Â </span>It is
|
---|
1022 |
|
---|
1023 | suggested that these techniques could be extended to the browsing of more
|
---|
1024 |
|
---|
1025 | general networks.</span></p>
|
---|
1026 |
|
---|
1027 |
|
---|
1028 |
|
---|
1029 |
|
---|
1030 |
|
---|
1031 |
|
---|
1032 |
|
---|
1033 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/6</span></p>
|
---|
1034 |
|
---|
1035 |
|
---|
1036 |
|
---|
1037 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Facilitating
|
---|
1038 |
|
---|
1039 | multiple copy/past operations</span></p>
|
---|
1040 |
|
---|
1041 |
|
---|
1042 |
|
---|
1043 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Mark <span
|
---|
1044 |
|
---|
1045 | class=SpellE>Apperley</span>, Jay Baker, Dale Fletcher, Bill Rogers</span></p>
|
---|
1046 |
|
---|
1047 |
|
---|
1048 |
|
---|
1049 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Copy and paste,
|
---|
1050 |
|
---|
1051 | or cut and paste, using a clipboard or paste buffer has long been the principle
|
---|
1052 |
|
---|
1053 | facility provided to users for transferring data between and within GUI
|
---|
1054 |
|
---|
1055 | applications.<span style='mso-spacerun:yes'>Â </span>We argue that this
|
---|
1056 |
|
---|
1057 | mechanism can be clumsy in circumstances where several pieces of information
|
---|
1058 |
|
---|
1059 | must be moved systematically.<span style='mso-spacerun:yes'>Â </span>In two
|
---|
1060 |
|
---|
1061 | situations - extraction of data fields from unstructured data found in a
|
---|
1062 |
|
---|
1063 | directed search process, and reorganisation of computer program source text -
|
---|
1064 |
|
---|
1065 | we present alternative, more natural, user interface facilities to make the
|
---|
1066 |
|
---|
1067 | task less onerous, and to provide improved visual feedback during the
|
---|
1068 |
|
---|
1069 | operation.</span></p>
|
---|
1070 |
|
---|
1071 |
|
---|
1072 |
|
---|
1073 |
|
---|
1074 |
|
---|
1075 |
|
---|
1076 |
|
---|
1077 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>For the data
|
---|
1078 |
|
---|
1079 | extraction task we introduce the Stretchable Selection Tool, a <span
|
---|
1080 |
|
---|
1081 | class=SpellE>semi</span>-transparent overlay augmenting the mouse pointer to
|
---|
1082 |
|
---|
1083 | automate paste operations and provide information to prompt the user.<span
|
---|
1084 |
|
---|
1085 | style='mso-spacerun:yes'>Â </span>We describe a prototype implementation that
|
---|
1086 |
|
---|
1087 | functions in a collaborative software environment, allowing users to <span
|
---|
1088 |
|
---|
1089 | class=SpellE>cooperate</span> on a multiple copy/paste operation.<span
|
---|
1090 |
|
---|
1091 | style='mso-spacerun:yes'>Â </span>For text reorganisation, we present an
|
---|
1092 |
|
---|
1093 | extension to <span class=SpellE>Emacs</span>, providing similar functionality,
|
---|
1094 |
|
---|
1095 | but without the collaborative features.</span></p>
|
---|
1096 |
|
---|
1097 |
|
---|
1098 |
|
---|
1099 |
|
---|
1100 |
|
---|
1101 |
|
---|
1102 |
|
---|
1103 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/7</span></p>
|
---|
1104 |
|
---|
1105 |
|
---|
1106 |
|
---|
1107 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Automating
|
---|
1108 |
|
---|
1109 | iterative tasks with programming by demonstration: a user evaluation</span></p>
|
---|
1110 |
|
---|
1111 |
|
---|
1112 |
|
---|
1113 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Gordon W. <span
|
---|
1114 |
|
---|
1115 | class=SpellE>Paynter</span>, Ian H. <span class=SpellE>Witten</span></span></p>
|
---|
1116 |
|
---|
1117 |
|
---|
1118 |
|
---|
1119 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Computer users
|
---|
1120 |
|
---|
1121 | often face iterative tasks that cannot be automated using the tools and
|
---|
1122 |
|
---|
1123 | aggregation techniques provided by their application program: they end up
|
---|
1124 |
|
---|
1125 | performing the iteration by hand, repeating user interface actions over and
|
---|
1126 |
|
---|
1127 | over again.<span style='mso-spacerun:yes'>Â </span>We have implemented an
|
---|
1128 |
|
---|
1129 | agent, called Familiar, that can be taught to perform iterative tasks using
|
---|
1130 |
|
---|
1131 | programming by demonstration (PBD).<span style='mso-spacerun:yes'>Â
|
---|
1132 |
|
---|
1133 | </span>Unlike other PBD systems, it is domain independent and works with
|
---|
1134 |
|
---|
1135 | unmodified, widely-used, applications in a popular operating system.<span
|
---|
1136 |
|
---|
1137 | style='mso-spacerun:yes'>Â </span>In a formal evaluation, we found that users
|
---|
1138 |
|
---|
1139 | quickly learned to use the agent to automate iterative tasks.<span
|
---|
1140 |
|
---|
1141 | style='mso-spacerun:yes'>Â </span>Generally, the participants preferred to use
|
---|
1142 |
|
---|
1143 | multiple selection where possible, but could and did use PBD in situations
|
---|
1144 |
|
---|
1145 | involving iteration over many commands, or when other techniques were
|
---|
1146 |
|
---|
1147 | unavailable.</span></p>
|
---|
1148 |
|
---|
1149 |
|
---|
1150 |
|
---|
1151 |
|
---|
1152 |
|
---|
1153 |
|
---|
1154 |
|
---|
1155 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/8</span></p>
|
---|
1156 |
|
---|
1157 |
|
---|
1158 |
|
---|
1159 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>A survey of
|
---|
1160 |
|
---|
1161 | software requirements specification practices in the New Zealand software
|
---|
1162 |
|
---|
1163 | industry</span></p>
|
---|
1164 |
|
---|
1165 |
|
---|
1166 |
|
---|
1167 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Lindsay Groves,
|
---|
1168 |
|
---|
1169 | Ray <span class=SpellE>Nickson</span>, Greg Reeve, Steve Reeves, Mark Utting</span></p>
|
---|
1170 |
|
---|
1171 |
|
---|
1172 |
|
---|
1173 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We report on the
|
---|
1174 |
|
---|
1175 | software development techniques used in the New Zealand software industry,
|
---|
1176 |
|
---|
1177 | paying particular attention to requirements gathering.<span
|
---|
1178 |
|
---|
1179 | style='mso-spacerun:yes'>Â </span>We surveyed a selection of software companies
|
---|
1180 |
|
---|
1181 | with a general questionnaire and then conducted in-depth interviews with four
|
---|
1182 |
|
---|
1183 | companies.<span style='mso-spacerun:yes'>Â </span>Our results show a wide
|
---|
1184 |
|
---|
1185 | variety in the kinds of companies undertaking software development, employing a
|
---|
1186 |
|
---|
1187 | wide range of software development techniques.<span style='mso-spacerun:yes'>Â
|
---|
1188 |
|
---|
1189 | </span>Although our data are not sufficiently detailed to draw statistically
|
---|
1190 |
|
---|
1191 | significant conclusions, it appears that larger software development groups
|
---|
1192 |
|
---|
1193 | typically have more well-defined software development processes, spend
|
---|
1194 |
|
---|
1195 | proportionally more time on requirements gathering, and follow more rigorous
|
---|
1196 |
|
---|
1197 | testing regimes.</span></p>
|
---|
1198 |
|
---|
1199 |
|
---|
1200 |
|
---|
1201 |
|
---|
1202 |
|
---|
1203 |
|
---|
1204 |
|
---|
1205 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/9</span></p>
|
---|
1206 |
|
---|
1207 |
|
---|
1208 |
|
---|
1209 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The LRU*WWW proxy
|
---|
1210 |
|
---|
1211 | cache document replacement algorithm</span></p>
|
---|
1212 |
|
---|
1213 |
|
---|
1214 |
|
---|
1215 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Chung-<span
|
---|
1216 |
|
---|
1217 | class=SpellE>yi</span> Chang, Tony <span class=SpellE>McGregor</span>, Geoffrey
|
---|
1218 |
|
---|
1219 | Holmes</span></p>
|
---|
1220 |
|
---|
1221 |
|
---|
1222 |
|
---|
1223 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Obtaining good
|
---|
1224 |
|
---|
1225 | performance from WWW proxy caches is critically dependent on the document
|
---|
1226 |
|
---|
1227 | replacement policy used by the proxy.<span style='mso-spacerun:yes'>Â
|
---|
1228 |
|
---|
1229 | </span>This paper validates the work of other authors by reproducing their
|
---|
1230 |
|
---|
1231 | studies of proxy cache document replacement algorithms.<span
|
---|
1232 |
|
---|
1233 | style='mso-spacerun:yes'>Â </span>From this basis a cross-trace study is
|
---|
1234 |
|
---|
1235 | mounted.<span style='mso-spacerun:yes'>Â </span>This demonstrates that the
|
---|
1236 |
|
---|
1237 | performance of most document replacement algorithms is dependent on the type of
|
---|
1238 |
|
---|
1239 | workload that they are presented with.<span style='mso-spacerun:yes'>Â
|
---|
1240 |
|
---|
1241 | </span>Finally we propose a new algorithm, LRU*, that consistently performs
|
---|
1242 |
|
---|
1243 | well across all our traces.</span></p>
|
---|
1244 |
|
---|
1245 |
|
---|
1246 |
|
---|
1247 |
|
---|
1248 |
|
---|
1249 |
|
---|
1250 |
|
---|
1251 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/10</span></p>
|
---|
1252 |
|
---|
1253 |
|
---|
1254 |
|
---|
1255 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Reduced-error
|
---|
1256 |
|
---|
1257 | pruning with significance tests</span></p>
|
---|
1258 |
|
---|
1259 |
|
---|
1260 |
|
---|
1261 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
1262 |
|
---|
1263 | lang=EN-GB>Eibe</span></span><span lang=EN-GB> Frank, Ian H. <span
|
---|
1264 |
|
---|
1265 | class=SpellE>Witten</span></span></p>
|
---|
1266 |
|
---|
1267 |
|
---|
1268 |
|
---|
1269 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>When building
|
---|
1270 |
|
---|
1271 | classification models, it is common practice to prune them to counter spurious
|
---|
1272 |
|
---|
1273 | effects of the training data: this often improves performance and reduces model
|
---|
1274 |
|
---|
1275 | size.<span style='mso-spacerun:yes'>Â </span>&quot;Reduced-error pruning&quot;
|
---|
1276 |
|
---|
1277 | is a fast pruning procedure for decision trees that is known to produce small
|
---|
1278 |
|
---|
1279 | and accurate trees.<span style='mso-spacerun:yes'>Â </span>Apart from the data
|
---|
1280 |
|
---|
1281 | from which the tree is grown, it uses an independent &quot;pruning&quot; set,
|
---|
1282 |
|
---|
1283 | and pruning decisions are based on the model's error rate on this fresh
|
---|
1284 |
|
---|
1285 | data.<span style='mso-spacerun:yes'>Â </span>Recently it has been observed that
|
---|
1286 |
|
---|
1287 | reduced-error pruning <span class=SpellE>overfits</span> the pruning data,
|
---|
1288 |
|
---|
1289 | producing unnecessarily large decision trees.<span style='mso-spacerun:yes'>Â
|
---|
1290 |
|
---|
1291 | </span>This paper investigates whether standard statistical significance tests
|
---|
1292 |
|
---|
1293 | can be used to counter this phenomenon.</span></p>
|
---|
1294 |
|
---|
1295 |
|
---|
1296 |
|
---|
1297 |
|
---|
1298 |
|
---|
1299 |
|
---|
1300 |
|
---|
1301 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The problem of <span
|
---|
1302 |
|
---|
1303 | class=SpellE>overfitting</span> to the pruning set highlights the need for
|
---|
1304 |
|
---|
1305 | significance testing.<span style='mso-spacerun:yes'>Â </span>We investigate two
|
---|
1306 |
|
---|
1307 | classes of test, &quot;parametric&quot; and &quot;non-parametric.&quot;<span
|
---|
1308 |
|
---|
1309 | style='mso-spacerun:yes'>Â </span>The standard chi-squared statistic can be
|
---|
1310 |
|
---|
1311 | used both in a parametric test and as the basis for a non-parametric
|
---|
1312 |
|
---|
1313 | permutation test.<span style='mso-spacerun:yes'>Â </span>In both cases it is
|
---|
1314 |
|
---|
1315 | necessary to select the significance level at which pruning is applied.<span
|
---|
1316 |
|
---|
1317 | style='mso-spacerun:yes'>Â </span>We show empirically that both versions of the
|
---|
1318 |
|
---|
1319 | chi-squared test perform equally well if their significance levels are adjusted
|
---|
1320 |
|
---|
1321 | appropriately.<span style='mso-spacerun:yes'>Â </span>Using a collection of
|
---|
1322 |
|
---|
1323 | standard <span class=SpellE>datasets</span>, we show that significance testing
|
---|
1324 |
|
---|
1325 | improves on standard reduced error pruning if the significance level is
|
---|
1326 |
|
---|
1327 | tailored to the particular <span class=SpellE>dataset</span> at hand using
|
---|
1328 |
|
---|
1329 | cross-validation, yielding consistently smaller trees that perform at least as
|
---|
1330 |
|
---|
1331 | well and sometimes better.</span></p>
|
---|
1332 |
|
---|
1333 |
|
---|
1334 |
|
---|
1335 |
|
---|
1336 |
|
---|
1337 |
|
---|
1338 |
|
---|
1339 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/11</span></p>
|
---|
1340 |
|
---|
1341 |
|
---|
1342 |
|
---|
1343 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
1344 |
|
---|
1345 | lang=EN-GB>Weka</span></span><span lang=EN-GB>: Practical machine learning
|
---|
1346 |
|
---|
1347 | tools and techniques with Java implementations</span></p>
|
---|
1348 |
|
---|
1349 |
|
---|
1350 |
|
---|
1351 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Ian H. <span
|
---|
1352 |
|
---|
1353 | class=SpellE>Witten</span>, <span class=SpellE>Eibe</span> Frank, Len <span
|
---|
1354 |
|
---|
1355 | class=SpellE>Trigg</span>, Mark Hall, Geoffrey Holmes, Sally Jo Cunningham</span></p>
|
---|
1356 |
|
---|
1357 |
|
---|
1358 |
|
---|
1359 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The Waikato
|
---|
1360 |
|
---|
1361 | Environment for Knowledge Analysis (Weka) is a comprehensive suite of Java
|
---|
1362 |
|
---|
1363 | class libraries that implement many state-of-the-art machine learning and data
|
---|
1364 |
|
---|
1365 | mining algorithms.<span style='mso-spacerun:yes'>Â </span><span class=SpellE>Weka</span>
|
---|
1366 |
|
---|
1367 | is freely available on the <span class=SpellE>World-Wide</span> Web and
|
---|
1368 |
|
---|
1369 | accompanies a new text on data mining [1] which documents and fully explains
|
---|
1370 |
|
---|
1371 | all the algorithms it contains.<span style='mso-spacerun:yes'>Â
|
---|
1372 |
|
---|
1373 | </span>Applications written using the <span class=SpellE>Weka</span> class
|
---|
1374 |
|
---|
1375 | libraries can be run on any computer with a Web browsing capability; this
|
---|
1376 |
|
---|
1377 | allows users to apply machine learning techniques to their own data regardless
|
---|
1378 |
|
---|
1379 | of computer platform.</span></p>
|
---|
1380 |
|
---|
1381 |
|
---|
1382 |
|
---|
1383 |
|
---|
1384 |
|
---|
1385 |
|
---|
1386 |
|
---|
1387 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/12</span></p>
|
---|
1388 |
|
---|
1389 |
|
---|
1390 |
|
---|
1391 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Pace Regression</span></p>
|
---|
1392 |
|
---|
1393 |
|
---|
1394 |
|
---|
1395 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Yong Wang, Ian
|
---|
1396 |
|
---|
1397 | H. <span class=SpellE>Witten</span></span></p>
|
---|
1398 |
|
---|
1399 |
|
---|
1400 |
|
---|
1401 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
1402 |
|
---|
1403 | articulates a new method of linear regression, âpace regressionâ, that
|
---|
1404 |
|
---|
1405 | addresses many drawbacks of standard regression reported in the
|
---|
1406 |
|
---|
1407 | literatureâparticularly the subset selection problem.<span
|
---|
1408 |
|
---|
1409 | style='mso-spacerun:yes'>Â </span>Pace regression improves on classical ordinary
|
---|
1410 |
|
---|
1411 | least squares (OLS) regression by evaluating the effect of each variable and
|
---|
1412 |
|
---|
1413 | using a clustering analysis to improve the statistical basis for estimating
|
---|
1414 |
|
---|
1415 | their contribution to the overall regression.<span style='mso-spacerun:yes'>Â
|
---|
1416 |
|
---|
1417 | </span>As well as outperforming OLS, it also outperformsâin a remarkably
|
---|
1418 |
|
---|
1419 | general senseâother linear <span class=SpellE>modeling</span> techniques in the
|
---|
1420 |
|
---|
1421 | literature, including subset selection procedures, which seek a reduction in
|
---|
1422 |
|
---|
1423 | dimensionality that falls out as a natural <span class=SpellE>byproduct</span>
|
---|
1424 |
|
---|
1425 | of pace regression.<span style='mso-spacerun:yes'>Â </span>The paper defines
|
---|
1426 |
|
---|
1427 | six procedures that share the fundamental idea of pace regression, all of which
|
---|
1428 |
|
---|
1429 | are theoretically justified in terms of asymptotic performance.<span
|
---|
1430 |
|
---|
1431 | style='mso-spacerun:yes'>Â </span>Experiments confirm the performance
|
---|
1432 |
|
---|
1433 | improvement over other techniques.</span></p>
|
---|
1434 |
|
---|
1435 |
|
---|
1436 |
|
---|
1437 |
|
---|
1438 |
|
---|
1439 |
|
---|
1440 |
|
---|
1441 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/13</span></p>
|
---|
1442 |
|
---|
1443 |
|
---|
1444 |
|
---|
1445 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>A
|
---|
1446 |
|
---|
1447 | compression-based algorithm for Chinese word segmentation</span></p>
|
---|
1448 |
|
---|
1449 |
|
---|
1450 |
|
---|
1451 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>W.J. <span
|
---|
1452 |
|
---|
1453 | class=SpellE>Teahan</span>, <span class=SpellE>Yingying</span> Wen, <span
|
---|
1454 |
|
---|
1455 | class=SpellE>Rodger</span> <span class=SpellE>McNab</span>, Ian H. <span
|
---|
1456 |
|
---|
1457 | class=SpellE>Witten</span></span></p>
|
---|
1458 |
|
---|
1459 |
|
---|
1460 |
|
---|
1461 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The Chinese
|
---|
1462 |
|
---|
1463 | language is written without using spaces or other word delimiters.<span
|
---|
1464 |
|
---|
1465 | style='mso-spacerun:yes'>Â </span>Although a text may be thought of as a
|
---|
1466 |
|
---|
1467 | corresponding sequence of words, there is considerable ambiguity in the
|
---|
1468 |
|
---|
1469 | placement of boundaries.<span style='mso-spacerun:yes'>Â </span>Interpreting a
|
---|
1470 |
|
---|
1471 | text as a sequence of words is beneficial for some information retrieval and
|
---|
1472 |
|
---|
1473 | storage tasks: for example, full-text search, word-based compression, and <span
|
---|
1474 |
|
---|
1475 | class=SpellE>keyphrase</span> extraction.</span></p>
|
---|
1476 |
|
---|
1477 |
|
---|
1478 |
|
---|
1479 |
|
---|
1480 |
|
---|
1481 |
|
---|
1482 |
|
---|
1483 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We describe a
|
---|
1484 |
|
---|
1485 | scheme that infers appropriate positions for word boundaries using an adaptive
|
---|
1486 |
|
---|
1487 | language model that is standard in text compression.<span
|
---|
1488 |
|
---|
1489 | style='mso-spacerun:yes'>Â </span>It is trained on a corpus of pre-segmented
|
---|
1490 |
|
---|
1491 | text, and when applied to new text, interpolates word boundaries so as to <span
|
---|
1492 |
|
---|
1493 | class=SpellE>maximize</span> the compression obtained.<span
|
---|
1494 |
|
---|
1495 | style='mso-spacerun:yes'>Â </span>This simple and general method performs well
|
---|
1496 |
|
---|
1497 | with respect to <span class=SpellE>specialized</span> schemes for Chinese
|
---|
1498 |
|
---|
1499 | language segmentation.</span></p>
|
---|
1500 |
|
---|
1501 |
|
---|
1502 |
|
---|
1503 |
|
---|
1504 |
|
---|
1505 |
|
---|
1506 |
|
---|
1507 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/14</span></p>
|
---|
1508 |
|
---|
1509 |
|
---|
1510 |
|
---|
1511 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Clustering with
|
---|
1512 |
|
---|
1513 | finite data from <span class=SpellE>semi</span>-parametric mixture
|
---|
1514 |
|
---|
1515 | distributions</span></p>
|
---|
1516 |
|
---|
1517 |
|
---|
1518 |
|
---|
1519 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Yong Wang, Ian
|
---|
1520 |
|
---|
1521 | H. <span class=SpellE>Witten</span></span></p>
|
---|
1522 |
|
---|
1523 |
|
---|
1524 |
|
---|
1525 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Existing
|
---|
1526 |
|
---|
1527 | clustering methods for the <span class=SpellE>semi</span>-parametric mixture
|
---|
1528 |
|
---|
1529 | distribution perform well as the volume of data increases.<span
|
---|
1530 |
|
---|
1531 | style='mso-spacerun:yes'>Â </span>However, they all suffer from a serious
|
---|
1532 |
|
---|
1533 | drawback in finite-data situations: small outlying groups of data points can be
|
---|
1534 |
|
---|
1535 | completely ignored in the clusters that are produced, no matter how far away
|
---|
1536 |
|
---|
1537 | they lie from the major clusters.<span style='mso-spacerun:yes'>Â </span>This
|
---|
1538 |
|
---|
1539 | can result in unbounded loss if the loss function is sensitive to the distance
|
---|
1540 |
|
---|
1541 | between clusters.</span></p>
|
---|
1542 |
|
---|
1543 |
|
---|
1544 |
|
---|
1545 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
1546 |
|
---|
1547 | proposes a new distance-based clustering method that overcomes the problem by
|
---|
1548 |
|
---|
1549 | avoiding global constraints.<span style='mso-spacerun:yes'>Â
|
---|
1550 |
|
---|
1551 | </span>Experimental results illustrate its superiority to existing methods when
|
---|
1552 |
|
---|
1553 | small clusters are present in finite data sets; they also suggest that it is
|
---|
1554 |
|
---|
1555 | more accurate and stable than other methods even when there are no small
|
---|
1556 |
|
---|
1557 | clusters.</span></p>
|
---|
1558 |
|
---|
1559 |
|
---|
1560 |
|
---|
1561 |
|
---|
1562 |
|
---|
1563 |
|
---|
1564 |
|
---|
1565 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/15</span></p>
|
---|
1566 |
|
---|
1567 |
|
---|
1568 |
|
---|
1569 |
|
---|
1570 |
|
---|
1571 |
|
---|
1572 |
|
---|
1573 |
|
---|
1574 |
|
---|
1575 |
|
---|
1576 |
|
---|
1577 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>99/16</span></p>
|
---|
1578 |
|
---|
1579 |
|
---|
1580 |
|
---|
1581 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The <span
|
---|
1582 |
|
---|
1583 | class=SpellE>Niupepa</span> Collection:<span style='mso-spacerun:yes'>Â </span>Opening
|
---|
1584 |
|
---|
1585 | the blinds on a window to the past</span></p>
|
---|
1586 |
|
---|
1587 |
|
---|
1588 |
|
---|
1589 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
1590 |
|
---|
1591 | lang=EN-GB>Te</span></span><span lang=EN-GB> <span class=SpellE>Taka</span> <span
|
---|
1592 |
|
---|
1593 | class=SpellE>Keegan</span>, Sally Jo Cunningham, Mark <span class=SpellE>Apperley</span></span></p>
|
---|
1594 |
|
---|
1595 |
|
---|
1596 |
|
---|
1597 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
1598 |
|
---|
1599 | describes the building of a digital library collection of historic
|
---|
1600 |
|
---|
1601 | newspapers.<span style='mso-spacerun:yes'>Â </span>The newspapers (<span
|
---|
1602 |
|
---|
1603 | class=SpellE><i style='mso-bidi-font-style:normal'>Niupepa</i></span> in <span
|
---|
1604 |
|
---|
1605 | class=SpellE>Maori</span>), which were published in New Zealand during the
|
---|
1606 |
|
---|
1607 | period 1842 to 1933, form a unique historical record of the <span class=SpellE>Maori</span>
|
---|
1608 |
|
---|
1609 | language, and of events from an historical perspective.<span
|
---|
1610 |
|
---|
1611 | style='mso-spacerun:yes'>Â </span>Images of these newspapers have been
|
---|
1612 |
|
---|
1613 | converted to digital form, electronic text extracted from these, and the
|
---|
1614 |
|
---|
1615 | collection is now being made available over the Internet as a part of the New
|
---|
1616 |
|
---|
1617 | Zealand Digital Library (NZDL) project at the University of Waikato.</span></p>
|
---|
1618 |
|
---|
1619 |
|
---|
1620 |
|
---|
1621 |
|
---|
1622 |
|
---|
1623 |
|
---|
1624 |
|
---|
1625 |
|
---|
1626 |
|
---|
1627 |
|
---|
1628 |
|
---|
1629 |
|
---|
1630 |
|
---|
1631 |
|
---|
1632 |
|
---|
1633 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/1</span></p>
|
---|
1634 |
|
---|
1635 |
|
---|
1636 |
|
---|
1637 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Boosting trees
|
---|
1638 |
|
---|
1639 | for cost-sensitive classifications</span></p>
|
---|
1640 |
|
---|
1641 |
|
---|
1642 |
|
---|
1643 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Kai <span
|
---|
1644 |
|
---|
1645 | class=SpellE>Ming</span> Ting, <span class=SpellE>Zijian</span> <span
|
---|
1646 |
|
---|
1647 | class=SpellE>Zheng</span></span></p>
|
---|
1648 |
|
---|
1649 |
|
---|
1650 |
|
---|
1651 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
1652 |
|
---|
1653 | explores two boosting techniques for cost-sensitive tree classification in the
|
---|
1654 |
|
---|
1655 | situation where misclassification costs change very often.<span
|
---|
1656 |
|
---|
1657 | style='mso-spacerun:yes'>Â </span>Ideally, one would like to have only one
|
---|
1658 |
|
---|
1659 | induction, and use the induced model for different misclassification
|
---|
1660 |
|
---|
1661 | costs.<span style='mso-spacerun:yes'>Â </span>Thus, it demands robustness of
|
---|
1662 |
|
---|
1663 | the induced model against cost changes.<span style='mso-spacerun:yes'>Â
|
---|
1664 |
|
---|
1665 | </span>Combining multiple trees gives robust predictions against this
|
---|
1666 |
|
---|
1667 | change.<span style='mso-spacerun:yes'>Â </span>We demonstrate that ordinary
|
---|
1668 |
|
---|
1669 | boosting combined with the minimum expected cost criterion to select the
|
---|
1670 |
|
---|
1671 | prediction class is a good solution under this situation.<span
|
---|
1672 |
|
---|
1673 | style='mso-spacerun:yes'>Â </span>We also introduce a variant of the ordinary
|
---|
1674 |
|
---|
1675 | boosting procedure which <span class=SpellE>utilizes</span> the cost
|
---|
1676 |
|
---|
1677 | information during training.<span style='mso-spacerun:yes'>Â </span>We show
|
---|
1678 |
|
---|
1679 | that the proposed technique performs better than the ordinary boosting in terms
|
---|
1680 |
|
---|
1681 | of misclassification cost.<span style='mso-spacerun:yes'>Â </span>However, this
|
---|
1682 |
|
---|
1683 | technique requires to induce a set of new trees every time the cost
|
---|
1684 |
|
---|
1685 | changes.<span style='mso-spacerun:yes'>Â </span>Our empirical investigation
|
---|
1686 |
|
---|
1687 | also reveals some interesting <span class=SpellE>behavior</span> of boosting
|
---|
1688 |
|
---|
1689 | decision trees for cost-sensitive classification.</span></p>
|
---|
1690 |
|
---|
1691 |
|
---|
1692 |
|
---|
1693 |
|
---|
1694 |
|
---|
1695 |
|
---|
1696 |
|
---|
1697 |
|
---|
1698 |
|
---|
1699 |
|
---|
1700 |
|
---|
1701 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/2</span></p>
|
---|
1702 |
|
---|
1703 |
|
---|
1704 |
|
---|
1705 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Generating
|
---|
1706 |
|
---|
1707 | accurate rule sets without global <span class=SpellE>optimization</span> </span></p>
|
---|
1708 |
|
---|
1709 |
|
---|
1710 |
|
---|
1711 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
1712 |
|
---|
1713 | lang=EN-GB>Eibe</span></span><span lang=EN-GB> Frank, Ian H. <span
|
---|
1714 |
|
---|
1715 | class=SpellE>Witten</span></span></p>
|
---|
1716 |
|
---|
1717 |
|
---|
1718 |
|
---|
1719 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The two dominant
|
---|
1720 |
|
---|
1721 | schemes for rule-learning, C4.5 and RIPPER, both operate in two stages.<span
|
---|
1722 |
|
---|
1723 | style='mso-spacerun:yes'>Â </span>First they induce an initial rule set and
|
---|
1724 |
|
---|
1725 | then they refine it using a rather complex <span class=SpellE>optimization</span>
|
---|
1726 |
|
---|
1727 | stage that discards (C4.5) or adjusts (RIPPER) individual rules to make them
|
---|
1728 |
|
---|
1729 | work better together.<span style='mso-spacerun:yes'>Â </span>In contrast, this
|
---|
1730 |
|
---|
1731 | paper shows how good rule sets can be learned one rule at a time, without any
|
---|
1732 |
|
---|
1733 | need for global <span class=SpellE>optimization</span>.<span
|
---|
1734 |
|
---|
1735 | style='mso-spacerun:yes'>Â </span>We present an algorithm for inferring rules
|
---|
1736 |
|
---|
1737 | by repeatedly generating partial decision trees, thus combining the two major
|
---|
1738 |
|
---|
1739 | paradigms for rule generation-creating rules from decision trees and the
|
---|
1740 |
|
---|
1741 | separate-and-conquer rule-learning technique.<span style='mso-spacerun:yes'>Â
|
---|
1742 |
|
---|
1743 | </span>The algorithm is straightforward and elegant: despite this, experiments
|
---|
1744 |
|
---|
1745 | on standard <span class=SpellE>datasets</span> show that it produces rule sets
|
---|
1746 |
|
---|
1747 | that are as accurate as and of similar size to those generated by C4.5, and
|
---|
1748 |
|
---|
1749 | more accurate than <span class=SpellE>RIPPER's</span>.<span
|
---|
1750 |
|
---|
1751 | style='mso-spacerun:yes'>Â </span>Moreover, it operates efficiently, and
|
---|
1752 |
|
---|
1753 | because it avoids <span class=SpellE>postprocessing</span>, does not suffer the
|
---|
1754 |
|
---|
1755 | extremely slow performance on pathological example sets for which the C4.5
|
---|
1756 |
|
---|
1757 | method has been <span class=SpellE>criticized</span>.</span></p>
|
---|
1758 |
|
---|
1759 |
|
---|
1760 |
|
---|
1761 |
|
---|
1762 |
|
---|
1763 |
|
---|
1764 |
|
---|
1765 |
|
---|
1766 |
|
---|
1767 |
|
---|
1768 |
|
---|
1769 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/3</span></p>
|
---|
1770 |
|
---|
1771 |
|
---|
1772 |
|
---|
1773 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
1774 |
|
---|
1775 | lang=EN-GB>VQuery</span></span><span lang=EN-GB>: a graphical user interface
|
---|
1776 |
|
---|
1777 | for Boolean query Specification and dynamic result preview</span></p>
|
---|
1778 |
|
---|
1779 |
|
---|
1780 |
|
---|
1781 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Steve Jones</span></p>
|
---|
1782 |
|
---|
1783 |
|
---|
1784 |
|
---|
1785 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Textual query
|
---|
1786 |
|
---|
1787 | languages based on Boolean logic are common amongst the search facilities of
|
---|
1788 |
|
---|
1789 | on-line information repositories.<span style='mso-spacerun:yes'>Â
|
---|
1790 |
|
---|
1791 | </span>However, there is evidence to suggest that the syntactic and semantic
|
---|
1792 |
|
---|
1793 | demands of such languages lead to user errors and adversely affect the time
|
---|
1794 |
|
---|
1795 | that it takes users to form queries.<span style='mso-spacerun:yes'>Â
|
---|
1796 |
|
---|
1797 | </span>Additionally, users are faced with user interfaces to these repositories
|
---|
1798 |
|
---|
1799 | which are unresponsive and uninformative, and consequently fail to support
|
---|
1800 |
|
---|
1801 | effective query refinement.<span style='mso-spacerun:yes'>Â </span>We suggest
|
---|
1802 |
|
---|
1803 | that graphical query languages, particularly Venn-like diagrams, provide a
|
---|
1804 |
|
---|
1805 | natural medium for Boolean query specification which overcomes the problems of
|
---|
1806 |
|
---|
1807 | textual query languages.<span style='mso-spacerun:yes'>Â </span>Also, dynamic
|
---|
1808 |
|
---|
1809 | result previews can be seamlessly integrated with graphical query specification
|
---|
1810 |
|
---|
1811 | to increase the effectiveness of query refinements.<span
|
---|
1812 |
|
---|
1813 | style='mso-spacerun:yes'>Â </span>We describe <span class=SpellE>VQuery</span>,
|
---|
1814 |
|
---|
1815 | a query interface to the New Zealand Digital Library which exploits querying by
|
---|
1816 |
|
---|
1817 | Venn diagrams and integrated query result previews.</span></p>
|
---|
1818 |
|
---|
1819 |
|
---|
1820 |
|
---|
1821 |
|
---|
1822 |
|
---|
1823 |
|
---|
1824 |
|
---|
1825 |
|
---|
1826 |
|
---|
1827 |
|
---|
1828 |
|
---|
1829 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/4</span></p>
|
---|
1830 |
|
---|
1831 |
|
---|
1832 |
|
---|
1833 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Revising
|
---|
1834 |
|
---|
1835 | &lt;I&gt;Z&lt;/I&gt;: semantics and logic</span></p>
|
---|
1836 |
|
---|
1837 |
|
---|
1838 |
|
---|
1839 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Martin C. <span
|
---|
1840 |
|
---|
1841 | class=SpellE>Henson</span>, Steve Reeves</span></p>
|
---|
1842 |
|
---|
1843 |
|
---|
1844 |
|
---|
1845 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We introduce a
|
---|
1846 |
|
---|
1847 | simple specification logic &lt;I&gt;Z&lt;/I&gt;c comprising a logic and
|
---|
1848 |
|
---|
1849 | semantics (in &lt;I&gt;ZF&lt;/I&gt; set theory).<span
|
---|
1850 |
|
---|
1851 | style='mso-spacerun:yes'>Â </span>We then provide an interpretation for (a
|
---|
1852 |
|
---|
1853 | rational reconstruction of) the specification language &lt;I&gt;Z&lt;/I&gt;
|
---|
1854 |
|
---|
1855 | within &lt;I&gt;Z&lt;/I&gt;c.<span style='mso-spacerun:yes'>Â </span>As a
|
---|
1856 |
|
---|
1857 | result we obtain a sound logic for &lt;I&gt;Z&lt;/I&gt;, including the schema
|
---|
1858 |
|
---|
1859 | calculus.<span style='mso-spacerun:yes'>Â </span>A consequence of our
|
---|
1860 |
|
---|
1861 | formalisation is a critique of a number of concepts used in
|
---|
1862 |
|
---|
1863 | &lt;I&gt;Z&lt;/I&gt;.<span style='mso-spacerun:yes'>Â </span>We demonstrate
|
---|
1864 |
|
---|
1865 | that the complications and confusions which these concepts introduce can be avoided
|
---|
1866 |
|
---|
1867 | without compromising <span class=SpellE>expressibility</span>.</span></p>
|
---|
1868 |
|
---|
1869 |
|
---|
1870 |
|
---|
1871 |
|
---|
1872 |
|
---|
1873 |
|
---|
1874 |
|
---|
1875 |
|
---|
1876 |
|
---|
1877 |
|
---|
1878 |
|
---|
1879 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/5</span></p>
|
---|
1880 |
|
---|
1881 |
|
---|
1882 |
|
---|
1883 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>A logic for the
|
---|
1884 |
|
---|
1885 | schema calculus</span></p>
|
---|
1886 |
|
---|
1887 |
|
---|
1888 |
|
---|
1889 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Martin C. <span
|
---|
1890 |
|
---|
1891 | class=SpellE>Henson</span>, Steve Reeves</span></p>
|
---|
1892 |
|
---|
1893 |
|
---|
1894 |
|
---|
1895 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>In this paper we
|
---|
1896 |
|
---|
1897 | introduce and investigate a logic for the schema calculus of
|
---|
1898 |
|
---|
1899 | &lt;I&gt;Z&lt;/I&gt;.<span style='mso-spacerun:yes'>Â </span>The schema
|
---|
1900 |
|
---|
1901 | calculus is arguably the reason for &lt;I&gt;Z&lt;/I&gt;âs popularity but so
|
---|
1902 |
|
---|
1903 | far no true calculus (a sound system of rules for reasoning about schema
|
---|
1904 |
|
---|
1905 | expressions) has been given.<span style='mso-spacerun:yes'>Â
|
---|
1906 |
|
---|
1907 | </span>Presentations thus far have either failed to provide a calculus (e.g.
|
---|
1908 |
|
---|
1909 | the draft standard [3]) or have fallen back on informal descriptions at a
|
---|
1910 |
|
---|
1911 | syntactic level (most text books e.g. [7[).<span style='mso-spacerun:yes'>Â
|
---|
1912 |
|
---|
1913 | </span>Once the calculus is established we introduce a derived <span
|
---|
1914 |
|
---|
1915 | class=SpellE>equational</span> logic which enables us to formalise properly the
|
---|
1916 |
|
---|
1917 | informal notations of schema expression equality to be found in the literature.</span></p>
|
---|
1918 |
|
---|
1919 |
|
---|
1920 |
|
---|
1921 |
|
---|
1922 |
|
---|
1923 |
|
---|
1924 |
|
---|
1925 |
|
---|
1926 |
|
---|
1927 |
|
---|
1928 |
|
---|
1929 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/6</span></p>
|
---|
1930 |
|
---|
1931 |
|
---|
1932 |
|
---|
1933 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>New foundations
|
---|
1934 |
|
---|
1935 | for &lt;I&gt;Z&lt;/I&gt;</span></p>
|
---|
1936 |
|
---|
1937 |
|
---|
1938 |
|
---|
1939 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Martin C. <span
|
---|
1940 |
|
---|
1941 | class=SpellE>Henson</span>, Steve Reeves</span></p>
|
---|
1942 |
|
---|
1943 |
|
---|
1944 |
|
---|
1945 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We provide a
|
---|
1946 |
|
---|
1947 | constructive and <span class=SpellE>intensional</span> interpretation for the
|
---|
1948 |
|
---|
1949 | specification language &lt;I&gt;Z&lt;/I&gt; in a theory of operations and kinds
|
---|
1950 |
|
---|
1951 | &lt;I&gt;T&lt;/I&gt;.<span style='mso-spacerun:yes'>Â </span>The motivation is
|
---|
1952 |
|
---|
1953 | to facilitate the development of an integrated approach to program
|
---|
1954 |
|
---|
1955 | construction.<span style='mso-spacerun:yes'>Â </span>We illustrate the new
|
---|
1956 |
|
---|
1957 | foundations for &lt;I&gt;Z&lt;/I&gt; with examples.</span></p>
|
---|
1958 |
|
---|
1959 |
|
---|
1960 |
|
---|
1961 |
|
---|
1962 |
|
---|
1963 |
|
---|
1964 |
|
---|
1965 |
|
---|
1966 |
|
---|
1967 |
|
---|
1968 |
|
---|
1969 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/7</span></p>
|
---|
1970 |
|
---|
1971 |
|
---|
1972 |
|
---|
1973 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Predicting apple
|
---|
1974 |
|
---|
1975 | bruising relationships using machine learning</span></p>
|
---|
1976 |
|
---|
1977 |
|
---|
1978 |
|
---|
1979 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>G. Holmes, S.J.
|
---|
1980 |
|
---|
1981 | Cunningham, B.T. <span class=SpellE>Dela</span> Rue, <span class=SpellE>A.F.</span>
|
---|
1982 |
|
---|
1983 | <span class=SpellE>Bollen</span></span></p>
|
---|
1984 |
|
---|
1985 |
|
---|
1986 |
|
---|
1987 | <p class=MsoBodyText><span lang=EN-US>Many models have been used to describe
|
---|
1988 |
|
---|
1989 | the influence of internal or external factors on apple bruising.<span
|
---|
1990 |
|
---|
1991 | style='mso-spacerun:yes'>Â </span>Few of these have addressed the application
|
---|
1992 |
|
---|
1993 | of derived relationships to the evaluation of commercial operations.<span
|
---|
1994 |
|
---|
1995 | style='mso-spacerun:yes'>Â </span>From an industry perspective, a model must
|
---|
1996 |
|
---|
1997 | enable fruit to be rejected on the basis of a commercially significant bruise
|
---|
1998 |
|
---|
1999 | and must also accurately quantify the effects of various combinations of input
|
---|
2000 |
|
---|
2001 | features (such as <span class=SpellE>cultivar</span>, maturity, size, and so
|
---|
2002 |
|
---|
2003 | on) on bruise prediction.<span style='mso-spacerun:yes'>Â </span>Input features
|
---|
2004 |
|
---|
2005 | must in turn have characteristics which are measurable commercially; for
|
---|
2006 |
|
---|
2007 | example, the measure of force should be impact energy rather than energy
|
---|
2008 |
|
---|
2009 | absorbed.<span style='mso-spacerun:yes'>Â </span>Further, as the commercial
|
---|
2010 |
|
---|
2011 | criteria for acceptable damage levels change, the model should be versatile
|
---|
2012 |
|
---|
2013 | enough to regenerate new bruise thresholds from existing data.</span></p>
|
---|
2014 |
|
---|
2015 |
|
---|
2016 |
|
---|
2017 |
|
---|
2018 |
|
---|
2019 |
|
---|
2020 |
|
---|
2021 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Machine learning
|
---|
2022 |
|
---|
2023 | is a burgeoning technology with a vast range of potential applications
|
---|
2024 |
|
---|
2025 | particularly in agriculture where large amounts of data can be readily
|
---|
2026 |
|
---|
2027 | collected [1].<span style='mso-spacerun:yes'>Â </span>The main advantage of
|
---|
2028 |
|
---|
2029 | using a machine learning method in an application is that the models built for
|
---|
2030 |
|
---|
2031 | prediction can be viewed and understood by the owner of the data who is in a
|
---|
2032 |
|
---|
2033 | position to determine the usefulness of the model, an essential component in a
|
---|
2034 |
|
---|
2035 | commercial environment.</span></p>
|
---|
2036 |
|
---|
2037 |
|
---|
2038 |
|
---|
2039 |
|
---|
2040 |
|
---|
2041 |
|
---|
2042 |
|
---|
2043 |
|
---|
2044 |
|
---|
2045 |
|
---|
2046 |
|
---|
2047 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/8</span></p>
|
---|
2048 |
|
---|
2049 |
|
---|
2050 |
|
---|
2051 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>An evaluation of
|
---|
2052 |
|
---|
2053 | passage-level indexing strategies for a technical report archive</span></p>
|
---|
2054 |
|
---|
2055 |
|
---|
2056 |
|
---|
2057 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Michael Williams</span></p>
|
---|
2058 |
|
---|
2059 |
|
---|
2060 |
|
---|
2061 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Past research
|
---|
2062 |
|
---|
2063 | has shown that using evidence from document passages rather than complete
|
---|
2064 |
|
---|
2065 | documents is an effective way of improving the precision of full-text database
|
---|
2066 |
|
---|
2067 | searches.<span style='mso-spacerun:yes'>Â </span>However, passage-level
|
---|
2068 |
|
---|
2069 | indexing has yet to be widely adopted for commercial or online databases.</span></p>
|
---|
2070 |
|
---|
2071 |
|
---|
2072 |
|
---|
2073 |
|
---|
2074 |
|
---|
2075 |
|
---|
2076 |
|
---|
2077 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
2078 |
|
---|
2079 | reports on experiments designed to test the efficacy of passage-level indexing
|
---|
2080 |
|
---|
2081 | with a particular collection of a full-text online database, the New Zealand
|
---|
2082 |
|
---|
2083 | Digital Library.<span style='mso-spacerun:yes'>Â </span>Discourse passages and
|
---|
2084 |
|
---|
2085 | word-window passages are used for the indexing process.<span
|
---|
2086 |
|
---|
2087 | style='mso-spacerun:yes'>Â </span>Both ranked and Boolean searching are used to
|
---|
2088 |
|
---|
2089 | test the resulting indexes.</span></p>
|
---|
2090 |
|
---|
2091 |
|
---|
2092 |
|
---|
2093 |
|
---|
2094 |
|
---|
2095 |
|
---|
2096 |
|
---|
2097 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Overlapping
|
---|
2098 |
|
---|
2099 | window passages are shown to offer the best retrieval performance with both
|
---|
2100 |
|
---|
2101 | ranked and Boolean queries.<span style='mso-spacerun:yes'>Â
|
---|
2102 |
|
---|
2103 | </span>Modifications may be necessary to the term weighting methodology in
|
---|
2104 |
|
---|
2105 | order to ensure optimal ranked query performance.</span></p>
|
---|
2106 |
|
---|
2107 |
|
---|
2108 |
|
---|
2109 |
|
---|
2110 |
|
---|
2111 |
|
---|
2112 |
|
---|
2113 |
|
---|
2114 |
|
---|
2115 |
|
---|
2116 |
|
---|
2117 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/9</span></p>
|
---|
2118 |
|
---|
2119 |
|
---|
2120 |
|
---|
2121 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Managing
|
---|
2122 |
|
---|
2123 | multiple collections, multiple languages, and multiple media in a distributed
|
---|
2124 |
|
---|
2125 | digital library</span></p>
|
---|
2126 |
|
---|
2127 |
|
---|
2128 |
|
---|
2129 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Ian H. <span
|
---|
2130 |
|
---|
2131 | class=SpellE>Witten</span>, <span class=SpellE>Rodger</span> <span
|
---|
2132 |
|
---|
2133 | class=SpellE>McNab</span>, Steve Jones, Sally Jo Cunningham, David Bainbridge,
|
---|
2134 |
|
---|
2135 | Mark <span class=SpellE>Apperley</span></span></p>
|
---|
2136 |
|
---|
2137 |
|
---|
2138 |
|
---|
2139 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Managing the <span
|
---|
2140 |
|
---|
2141 | class=SpellE>organizational</span> and software complexity of a comprehensive
|
---|
2142 |
|
---|
2143 | digital library presents a significant challenge.<span
|
---|
2144 |
|
---|
2145 | style='mso-spacerun:yes'>Â </span>Different library collections each have their
|
---|
2146 |
|
---|
2147 | own distinctive features.<span style='mso-spacerun:yes'>Â </span>Different
|
---|
2148 |
|
---|
2149 | presentation languages have structural implications such as left-to-right
|
---|
2150 |
|
---|
2151 | writing order and text-only interfaces for the visually impaired.<span
|
---|
2152 |
|
---|
2153 | style='mso-spacerun:yes'>Â </span>Different media involve different file
|
---|
2154 |
|
---|
2155 | formats, and-more importantly-radically different search strategies are
|
---|
2156 |
|
---|
2157 | required for non-textual media.<span style='mso-spacerun:yes'>Â </span>In a
|
---|
2158 |
|
---|
2159 | distributed library, new collections can appear asynchronously on servers in
|
---|
2160 |
|
---|
2161 | different parts of the world.<span style='mso-spacerun:yes'>Â </span>And as
|
---|
2162 |
|
---|
2163 | searching interfaces mature from the command-line era exemplified by current
|
---|
2164 |
|
---|
2165 | Web search engines into the age of reactive visual interfaces, experimental new
|
---|
2166 |
|
---|
2167 | interfaces must be developed, supported, and tested.<span
|
---|
2168 |
|
---|
2169 | style='mso-spacerun:yes'>Â </span>This paper describes our experience, gained
|
---|
2170 |
|
---|
2171 | from operating a substantial digital library service over several years, in
|
---|
2172 |
|
---|
2173 | solving these problems by designing an appropriate software architecture.</span></p>
|
---|
2174 |
|
---|
2175 |
|
---|
2176 |
|
---|
2177 |
|
---|
2178 |
|
---|
2179 |
|
---|
2180 |
|
---|
2181 |
|
---|
2182 |
|
---|
2183 |
|
---|
2184 |
|
---|
2185 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/10</span></p>
|
---|
2186 |
|
---|
2187 |
|
---|
2188 |
|
---|
2189 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Experiences with
|
---|
2190 |
|
---|
2191 | a weighted decision tree learner</span></p>
|
---|
2192 |
|
---|
2193 |
|
---|
2194 |
|
---|
2195 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>John G. <span
|
---|
2196 |
|
---|
2197 | class=SpellE>Cleary</span>, Leonard E. <span class=SpellE>Trigg</span></span></p>
|
---|
2198 |
|
---|
2199 |
|
---|
2200 |
|
---|
2201 | <p class=MsoBodyText><span lang=EN-US>Machine learning algorithms for inferring
|
---|
2202 |
|
---|
2203 | decision trees typically choose a single âbestâ tree to describe the training
|
---|
2204 |
|
---|
2205 | data.<span style='mso-spacerun:yes'>Â </span>Recent research has shown that
|
---|
2206 |
|
---|
2207 | classification performance can be significantly improved by voting predictions
|
---|
2208 |
|
---|
2209 | of multiple, independently produced decision trees.<span
|
---|
2210 |
|
---|
2211 | style='mso-spacerun:yes'>Â </span>This paper describes an algorithm, OB1, that
|
---|
2212 |
|
---|
2213 | makes a weighted sum over many possible models.<span style='mso-spacerun:yes'>Â
|
---|
2214 |
|
---|
2215 | </span>We describe one instance of OB1, that includes &lt;I&gt;all&lt;/I&gt;
|
---|
2216 |
|
---|
2217 | possible decision trees as well as naïve <span class=SpellE>Bayesian</span>
|
---|
2218 |
|
---|
2219 | models.<span style='mso-spacerun:yes'>Â </span>OB1 is compared with a number of
|
---|
2220 |
|
---|
2221 | other decision tree and instance based learning <span class=SpellE>alogrithms</span>
|
---|
2222 |
|
---|
2223 | on some of the data sets from the UCI repository.<span
|
---|
2224 |
|
---|
2225 | style='mso-spacerun:yes'>Â </span>Both an information gain and an accuracy
|
---|
2226 |
|
---|
2227 | measure are used for the comparison.<span style='mso-spacerun:yes'>Â </span>On
|
---|
2228 |
|
---|
2229 | the information gain measure OB1 performs significantly better than all the
|
---|
2230 |
|
---|
2231 | other algorithms.<span style='mso-spacerun:yes'>Â </span>On the accuracy
|
---|
2232 |
|
---|
2233 | measure it is significantly better than all the algorithms except naïve <span
|
---|
2234 |
|
---|
2235 | class=SpellE>Bayes</span> which performs comparably to OB1.</span></p>
|
---|
2236 |
|
---|
2237 |
|
---|
2238 |
|
---|
2239 |
|
---|
2240 |
|
---|
2241 |
|
---|
2242 |
|
---|
2243 |
|
---|
2244 |
|
---|
2245 |
|
---|
2246 |
|
---|
2247 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/11</span></p>
|
---|
2248 |
|
---|
2249 |
|
---|
2250 |
|
---|
2251 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>An entropy gain
|
---|
2252 |
|
---|
2253 | measure of numeric prediction performance</span></p>
|
---|
2254 |
|
---|
2255 |
|
---|
2256 |
|
---|
2257 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Leonard <span
|
---|
2258 |
|
---|
2259 | class=SpellE>Trigg</span></span></p>
|
---|
2260 |
|
---|
2261 |
|
---|
2262 |
|
---|
2263 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Categorical
|
---|
2264 |
|
---|
2265 | classifier performance is typically evaluated with respect to error rate,
|
---|
2266 |
|
---|
2267 | expressed as a percentage of test instances that were not correctly
|
---|
2268 |
|
---|
2269 | classified.<span style='mso-spacerun:yes'>Â </span>When a classifier produces
|
---|
2270 |
|
---|
2271 | multiple classifications for a test instance, the prediction is counted as
|
---|
2272 |
|
---|
2273 | incorrect (even if the correct class was one of the predictions).<span
|
---|
2274 |
|
---|
2275 | style='mso-spacerun:yes'>Â </span>Although commonly used in the literature,
|
---|
2276 |
|
---|
2277 | error rate is a coarse measure of classifier performance, as it is based only
|
---|
2278 |
|
---|
2279 | on a single prediction offered for a test instance.<span
|
---|
2280 |
|
---|
2281 | style='mso-spacerun:yes'>Â </span>Since many classifiers can produce a class
|
---|
2282 |
|
---|
2283 | distribution as a prediction, we should use this to provide a better measure of
|
---|
2284 |
|
---|
2285 | how much information the classifier is extracting from the domain.</span></p>
|
---|
2286 |
|
---|
2287 |
|
---|
2288 |
|
---|
2289 |
|
---|
2290 |
|
---|
2291 |
|
---|
2292 |
|
---|
2293 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Numeric
|
---|
2294 |
|
---|
2295 | classifiers are a relatively new development in machine learning, and as such
|
---|
2296 |
|
---|
2297 | there is no single performance measure that has become standard.<span
|
---|
2298 |
|
---|
2299 | style='mso-spacerun:yes'>Â </span>Typically these machine learning schemes
|
---|
2300 |
|
---|
2301 | predict a single real number for each test instance, and the error between the
|
---|
2302 |
|
---|
2303 | predicted and actual value is used to calculate a myriad of performance
|
---|
2304 |
|
---|
2305 | measures such as correlation coefficient, root mean squared error, mean
|
---|
2306 |
|
---|
2307 | absolute error, relative absolute error, and root relative squared error.<span
|
---|
2308 |
|
---|
2309 | style='mso-spacerun:yes'>Â </span>With so many performance measures it is
|
---|
2310 |
|
---|
2311 | difficult to establish an overall performance evaluation.</span></p>
|
---|
2312 |
|
---|
2313 |
|
---|
2314 |
|
---|
2315 |
|
---|
2316 |
|
---|
2317 |
|
---|
2318 |
|
---|
2319 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The next section
|
---|
2320 |
|
---|
2321 | describes a performance measure for machine learning schemes that attempts to
|
---|
2322 |
|
---|
2323 | overcome the problems with current measures.<span style='mso-spacerun:yes'>Â
|
---|
2324 |
|
---|
2325 | </span>In addition, the same evaluation measure is used for categorical and
|
---|
2326 |
|
---|
2327 | numeric classifier.</span></p>
|
---|
2328 |
|
---|
2329 |
|
---|
2330 |
|
---|
2331 |
|
---|
2332 |
|
---|
2333 |
|
---|
2334 |
|
---|
2335 |
|
---|
2336 |
|
---|
2337 |
|
---|
2338 |
|
---|
2339 |
|
---|
2340 |
|
---|
2341 |
|
---|
2342 |
|
---|
2343 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/12</span></p>
|
---|
2344 |
|
---|
2345 |
|
---|
2346 |
|
---|
2347 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Proceedings of
|
---|
2348 |
|
---|
2349 | CBISE â98 CaiSE*98 Workshop on Component Based Information Systems Engineering</span></p>
|
---|
2350 |
|
---|
2351 |
|
---|
2352 |
|
---|
2353 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Edited by John <span
|
---|
2354 |
|
---|
2355 | class=SpellE>Grundy</span></span></p>
|
---|
2356 |
|
---|
2357 |
|
---|
2358 |
|
---|
2359 | <p class=MsoBodyText><span lang=EN-US>Component-based information systems
|
---|
2360 |
|
---|
2361 | development is an area of research and practice of increasing importance.<span
|
---|
2362 |
|
---|
2363 | style='mso-spacerun:yes'>Â </span>Information Systems developers have <span
|
---|
2364 |
|
---|
2365 | class=SpellE>realised</span> that traditional approaches to IS engineering
|
---|
2366 |
|
---|
2367 | produce monolithic, difficult to maintain, difficult to reuse systems.<span
|
---|
2368 |
|
---|
2369 | style='mso-spacerun:yes'>Â </span>In contrast, the use of software components,
|
---|
2370 |
|
---|
2371 | which embody data, functionality and well-specified and understood interfaces,
|
---|
2372 |
|
---|
2373 | makes interoperable, distributed and highly reusable IS components
|
---|
2374 |
|
---|
2375 | feasible.<span style='mso-spacerun:yes'>Â </span>Component-based approaches to
|
---|
2376 |
|
---|
2377 | IS engineering can be used at strategic and <span class=SpellE>organisational</span>
|
---|
2378 |
|
---|
2379 | levels, to model business processes and whole IS architectures, in development
|
---|
2380 |
|
---|
2381 | methods which <span class=SpellE>utilise</span> component-based models during
|
---|
2382 |
|
---|
2383 | analysis and design, and in system implementation.<span
|
---|
2384 |
|
---|
2385 | style='mso-spacerun:yes'>Â </span>Reusable components can allow end users to
|
---|
2386 |
|
---|
2387 | compose and configure their own Information Systems, possibly from a range of
|
---|
2388 |
|
---|
2389 | suppliers, and to more tightly couple their <span class=SpellE>organisational</span>
|
---|
2390 |
|
---|
2391 | <span class=SpellE>workflows</span> with their IS support.</span></p>
|
---|
2392 |
|
---|
2393 |
|
---|
2394 |
|
---|
2395 |
|
---|
2396 |
|
---|
2397 |
|
---|
2398 |
|
---|
2399 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This workshop
|
---|
2400 |
|
---|
2401 | proceedings contains a range of papers addressing one or more of the above
|
---|
2402 |
|
---|
2403 | issues relating to the use of component models for IS development.<span
|
---|
2404 |
|
---|
2405 | style='mso-spacerun:yes'>Â </span>All of these papers were refereed by at least
|
---|
2406 |
|
---|
2407 | two members of an international workshop committee comprising industry and
|
---|
2408 |
|
---|
2409 | academic researchers and users of component technologies.<span
|
---|
2410 |
|
---|
2411 | style='mso-spacerun:yes'>Â </span>Strategic uses of components are addressed in
|
---|
2412 |
|
---|
2413 | the first three papers, while the following three address uses of components for
|
---|
2414 |
|
---|
2415 | systems design and workflow management.<span style='mso-spacerun:yes'>Â
|
---|
2416 |
|
---|
2417 | </span>Systems development using components, and the provision of environments
|
---|
2418 |
|
---|
2419 | for component management are addressed in the following group of five
|
---|
2420 |
|
---|
2421 | papers.<span style='mso-spacerun:yes'>Â </span>The last three papers in this
|
---|
2422 |
|
---|
2423 | proceedings address component management and analysis techniques.</span></p>
|
---|
2424 |
|
---|
2425 |
|
---|
2426 |
|
---|
2427 |
|
---|
2428 |
|
---|
2429 |
|
---|
2430 |
|
---|
2431 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>All of these
|
---|
2432 |
|
---|
2433 | papers provide new insights into the many<span style='mso-spacerun:yes'>Â
|
---|
2434 |
|
---|
2435 | </span>varied uses of component technology for IS engineering.<span
|
---|
2436 |
|
---|
2437 | style='mso-spacerun:yes'>Â </span>I hope you find them as interesting and
|
---|
2438 |
|
---|
2439 | useful as I have when collating this proceedings and organising the workshop.</span></p>
|
---|
2440 |
|
---|
2441 |
|
---|
2442 |
|
---|
2443 |
|
---|
2444 |
|
---|
2445 |
|
---|
2446 |
|
---|
2447 |
|
---|
2448 |
|
---|
2449 |
|
---|
2450 |
|
---|
2451 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/13</span></p>
|
---|
2452 |
|
---|
2453 |
|
---|
2454 |
|
---|
2455 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>An analysis of
|
---|
2456 |
|
---|
2457 | usage of a digital library</span></p>
|
---|
2458 |
|
---|
2459 |
|
---|
2460 |
|
---|
2461 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Steve Jones,
|
---|
2462 |
|
---|
2463 | Sally Jo Cunningham, <span class=SpellE>Rodger</span> <span class=SpellE>McNab</span></span></p>
|
---|
2464 |
|
---|
2465 |
|
---|
2466 |
|
---|
2467 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>As experimental
|
---|
2468 |
|
---|
2469 | digital library <span class=SpellE>testbeds</span> gain wider acceptance and
|
---|
2470 |
|
---|
2471 | develop significant user bases, it becomes important to investigate the ways in
|
---|
2472 |
|
---|
2473 | which users interact with the systems in practice.<span
|
---|
2474 |
|
---|
2475 | style='mso-spacerun:yes'>Â </span>Transaction logs are one source of usage
|
---|
2476 |
|
---|
2477 | information, and the information on user behaviour can be culled from them both
|
---|
2478 |
|
---|
2479 | automatically (through calculation of summary statistics) and manually (by
|
---|
2480 |
|
---|
2481 | examining query strings for semantic clues on search motivations and searching
|
---|
2482 |
|
---|
2483 | strategy).<span style='mso-spacerun:yes'>Â </span>We conduct a transaction log
|
---|
2484 |
|
---|
2485 | analysis on user activity in the Computer Science Technical Reports Collection
|
---|
2486 |
|
---|
2487 | of the New Zealand Digital Library, and report insights gained and identify
|
---|
2488 |
|
---|
2489 | resulting search interface design issues.</span></p>
|
---|
2490 |
|
---|
2491 |
|
---|
2492 |
|
---|
2493 |
|
---|
2494 |
|
---|
2495 |
|
---|
2496 |
|
---|
2497 |
|
---|
2498 |
|
---|
2499 |
|
---|
2500 |
|
---|
2501 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/14</span></p>
|
---|
2502 |
|
---|
2503 |
|
---|
2504 |
|
---|
2505 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Measuring ATM
|
---|
2506 |
|
---|
2507 | traffic: final report for New Zealand Telecom</span></p>
|
---|
2508 |
|
---|
2509 |
|
---|
2510 |
|
---|
2511 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>John <span
|
---|
2512 |
|
---|
2513 | class=SpellE>Cleary</span>, Ian Graham, Murray Pearson, Tony <span
|
---|
2514 |
|
---|
2515 | class=SpellE>McGregor</span></span></p>
|
---|
2516 |
|
---|
2517 |
|
---|
2518 |
|
---|
2519 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The report
|
---|
2520 |
|
---|
2521 | describes the development of a low-cost ATM monitoring system, hosted by a
|
---|
2522 |
|
---|
2523 | standard PC.<span style='mso-spacerun:yes'>Â </span>The monitor can be used
|
---|
2524 |
|
---|
2525 | remotely returning information on ATM traffic flows to a central site.<span
|
---|
2526 |
|
---|
2527 | style='mso-spacerun:yes'>Â </span>The monitor is interfaces to a GPS timing
|
---|
2528 |
|
---|
2529 | receiver, which provides an absolute time accuracy of better than 1 <span
|
---|
2530 |
|
---|
2531 | class=SpellE>usec</span>.<span style='mso-spacerun:yes'>Â </span>By monitoring
|
---|
2532 |
|
---|
2533 | the same traffic flow at different points in a network it is possible to
|
---|
2534 |
|
---|
2535 | measure cell delay and delay variation in real time, and with existing
|
---|
2536 |
|
---|
2537 | traffic.<span style='mso-spacerun:yes'>Â </span>The monitoring system
|
---|
2538 |
|
---|
2539 | characterises cells by a CRC calculated over the cell payload, thus special
|
---|
2540 |
|
---|
2541 | measurement cells are not required.<span style='mso-spacerun:yes'>Â
|
---|
2542 |
|
---|
2543 | </span>Delays in both local area and wide-area networks have been measured
|
---|
2544 |
|
---|
2545 | using this system.<span style='mso-spacerun:yes'>Â </span>It is possible to
|
---|
2546 |
|
---|
2547 | measure delay in a network that is not end-to-end ATM, as long as some cells
|
---|
2548 |
|
---|
2549 | remain identical at the entry and exit points.<span style='mso-spacerun:yes'>Â
|
---|
2550 |
|
---|
2551 | </span>Examples are given of traffic and delay measurements in both wide and
|
---|
2552 |
|
---|
2553 | local area network systems, including delays measured over the Internet from
|
---|
2554 |
|
---|
2555 | Canada to New Zealand.</span></p>
|
---|
2556 |
|
---|
2557 |
|
---|
2558 |
|
---|
2559 |
|
---|
2560 |
|
---|
2561 |
|
---|
2562 |
|
---|
2563 |
|
---|
2564 |
|
---|
2565 |
|
---|
2566 |
|
---|
2567 |
|
---|
2568 |
|
---|
2569 |
|
---|
2570 |
|
---|
2571 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/15</span></p>
|
---|
2572 |
|
---|
2573 |
|
---|
2574 |
|
---|
2575 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Despite its
|
---|
2576 |
|
---|
2577 | simplicity, the naïve <span class=SpellE>Bayes</span> learning scheme performs
|
---|
2578 |
|
---|
2579 | well on most classification tasks, and is often significantly more accurate
|
---|
2580 |
|
---|
2581 | than more sophisticated methods.<span style='mso-spacerun:yes'>Â
|
---|
2582 |
|
---|
2583 | </span>Although the probability estimates that it produces can be inaccurate,
|
---|
2584 |
|
---|
2585 | it often assigns maximum probability to the correct class.<span
|
---|
2586 |
|
---|
2587 | style='mso-spacerun:yes'>Â </span>This suggests that its good performance might
|
---|
2588 |
|
---|
2589 | be restricted to situations where the output is categorical.<span
|
---|
2590 |
|
---|
2591 | style='mso-spacerun:yes'>Â </span>It is therefore interesting to see how it
|
---|
2592 |
|
---|
2593 | performs in domains where the predicted value is numeric, because in this case,
|
---|
2594 |
|
---|
2595 | predictions are more sensitive to inaccurate probability estimates.&lt;P&gt;</span></p>
|
---|
2596 |
|
---|
2597 |
|
---|
2598 |
|
---|
2599 |
|
---|
2600 |
|
---|
2601 |
|
---|
2602 |
|
---|
2603 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper shows
|
---|
2604 |
|
---|
2605 | how to apply the naïve <span class=SpellE>Bayes</span> methodology to numeric
|
---|
2606 |
|
---|
2607 | prediction (i.e. regression) tasks, and compares it to linear regression,
|
---|
2608 |
|
---|
2609 | instance-based learning, and a method that produces âmodel treesâ-decision
|
---|
2610 |
|
---|
2611 | trees with linear regression functions at the leaves.<span
|
---|
2612 |
|
---|
2613 | style='mso-spacerun:yes'>Â </span>Although we exhibit an artificial <span
|
---|
2614 |
|
---|
2615 | class=SpellE>dataset</span> for which naïve <span class=SpellE>Bayes</span> is
|
---|
2616 |
|
---|
2617 | the method of choice, on real-world <span class=SpellE>datasets</span> it is
|
---|
2618 |
|
---|
2619 | almost uniformly worse than model trees.<span style='mso-spacerun:yes'>Â
|
---|
2620 |
|
---|
2621 | </span>The comparison with linear regression depends on the error measure: for
|
---|
2622 |
|
---|
2623 | one measure naïve <span class=SpellE>Bayes</span> performs similarly, for
|
---|
2624 |
|
---|
2625 | another it is worse.<span style='mso-spacerun:yes'>Â </span>Compared to
|
---|
2626 |
|
---|
2627 | instance-based learning, it performs similarly with respect to both
|
---|
2628 |
|
---|
2629 | measures.<span style='mso-spacerun:yes'>Â </span>These results indicate that
|
---|
2630 |
|
---|
2631 | the simplistic statistical assumption that naïve <span class=SpellE>Bayes</span>
|
---|
2632 |
|
---|
2633 | makes is indeed more restrictive for regression than for classification.</span></p>
|
---|
2634 |
|
---|
2635 |
|
---|
2636 |
|
---|
2637 |
|
---|
2638 |
|
---|
2639 |
|
---|
2640 |
|
---|
2641 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/16</span></p>
|
---|
2642 |
|
---|
2643 |
|
---|
2644 |
|
---|
2645 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Link as you
|
---|
2646 |
|
---|
2647 | type: using key phrases for automated dynamic link generation</span></p>
|
---|
2648 |
|
---|
2649 |
|
---|
2650 |
|
---|
2651 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Steve Jones</span></p>
|
---|
2652 |
|
---|
2653 |
|
---|
2654 |
|
---|
2655 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>When documents
|
---|
2656 |
|
---|
2657 | are collected together from diverse sources they are unlikely to contain useful
|
---|
2658 |
|
---|
2659 | hypertext links to support browsing amongst them.<span
|
---|
2660 |
|
---|
2661 | style='mso-spacerun:yes'>Â </span>For large collections of thousands of
|
---|
2662 |
|
---|
2663 | documents it is prohibitively resource intensive to manually insert links into
|
---|
2664 |
|
---|
2665 | each document.<span style='mso-spacerun:yes'>Â </span>Users of such collections
|
---|
2666 |
|
---|
2667 | may wish to relate documents within them to text that they are themselves
|
---|
2668 |
|
---|
2669 | generating.<span style='mso-spacerun:yes'>Â </span>This process, often
|
---|
2670 |
|
---|
2671 | involving keyword searching, distracts from the authoring process and results
|
---|
2672 |
|
---|
2673 | in material related to query terms but not necessarily to the authorâs
|
---|
2674 |
|
---|
2675 | document.<span style='mso-spacerun:yes'>Â </span>Query terms that are effective
|
---|
2676 |
|
---|
2677 | in one collection might not be so in another.<span style='mso-spacerun:yes'>Â
|
---|
2678 |
|
---|
2679 | </span>We have developed <span class=SpellE>Phrasier</span>, a system that
|
---|
2680 |
|
---|
2681 | integrates authoring (of text and hyperlinks), browsing, querying and reading
|
---|
2682 |
|
---|
2683 | in support of information retrieval activities.<span style='mso-spacerun:yes'>Â
|
---|
2684 |
|
---|
2685 | </span><span class=SpellE>Phrasier</span> exploits key phrases which are
|
---|
2686 |
|
---|
2687 | automatically extracted from documents in a collection, and uses them as link
|
---|
2688 |
|
---|
2689 | anchors and to identify candidate destinations for hyperlinks.<span
|
---|
2690 |
|
---|
2691 | style='mso-spacerun:yes'>Â </span>This system suggests links into existing
|
---|
2692 |
|
---|
2693 | collections for purposes of authoring and retrieval of related information,
|
---|
2694 |
|
---|
2695 | creates links between documents in a collection and provides supportive
|
---|
2696 |
|
---|
2697 | document and link overviews.</span></p>
|
---|
2698 |
|
---|
2699 |
|
---|
2700 |
|
---|
2701 |
|
---|
2702 |
|
---|
2703 |
|
---|
2704 |
|
---|
2705 |
|
---|
2706 |
|
---|
2707 |
|
---|
2708 |
|
---|
2709 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/17</span></p>
|
---|
2710 |
|
---|
2711 |
|
---|
2712 |
|
---|
2713 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Melody based
|
---|
2714 |
|
---|
2715 | tune retrieval over the World Wide Web</span></p>
|
---|
2716 |
|
---|
2717 |
|
---|
2718 |
|
---|
2719 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>David
|
---|
2720 |
|
---|
2721 | Bainbridge, <span class=SpellE>Rodger</span> J. <span class=SpellE>McNab</span>,
|
---|
2722 |
|
---|
2723 | Lloyd A. Smith</span></p>
|
---|
2724 |
|
---|
2725 |
|
---|
2726 |
|
---|
2727 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>In this paper we
|
---|
2728 |
|
---|
2729 | describe the steps taken to develop a Web-based version of an existing
|
---|
2730 |
|
---|
2731 | stand-alone, single-user digital library application for <span class=SpellE>melodical</span>
|
---|
2732 |
|
---|
2733 | searching of a collection of music.<span style='mso-spacerun:yes'>Â </span>For
|
---|
2734 |
|
---|
2735 | the three key components: input, searching, and output, we assess the
|
---|
2736 |
|
---|
2737 | suitability of various Web-based strategies that deal with the now distributed
|
---|
2738 |
|
---|
2739 | software architecture and explain the decisions we made.<span
|
---|
2740 |
|
---|
2741 | style='mso-spacerun:yes'>Â </span>The resulting melody indexing service, known
|
---|
2742 |
|
---|
2743 | as MELDEX, has been in operation for one year, and the feed-back we have
|
---|
2744 |
|
---|
2745 | received has been <span class=SpellE>favorable</span>.</span></p>
|
---|
2746 |
|
---|
2747 |
|
---|
2748 |
|
---|
2749 |
|
---|
2750 |
|
---|
2751 |
|
---|
2752 |
|
---|
2753 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>98/18</span></p>
|
---|
2754 |
|
---|
2755 |
|
---|
2756 |
|
---|
2757 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Making oral
|
---|
2758 |
|
---|
2759 | history accessible over the World Wide Web</span></p>
|
---|
2760 |
|
---|
2761 |
|
---|
2762 |
|
---|
2763 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>David
|
---|
2764 |
|
---|
2765 | Bainbridge, Sally Jo Cunningham</span></p>
|
---|
2766 |
|
---|
2767 |
|
---|
2768 |
|
---|
2769 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We describe a
|
---|
2770 |
|
---|
2771 | multimedia, WWW-based oral history collection constructed from off-the-shelf or
|
---|
2772 |
|
---|
2773 | publicly available software.<span style='mso-spacerun:yes'>Â </span>The source
|
---|
2774 |
|
---|
2775 | materials for the collection include audio tapes of interviews and summary
|
---|
2776 |
|
---|
2777 | transcripts of each interview, as well as photographs illustrating episodes
|
---|
2778 |
|
---|
2779 | mentioned in the tapes.<span style='mso-spacerun:yes'>Â </span>Sections of the
|
---|
2780 |
|
---|
2781 | transcripts are manually matched to associated segments of the tapes, and the
|
---|
2782 |
|
---|
2783 | tapes are <span class=SpellE>digitized</span>.<span style='mso-spacerun:yes'>Â
|
---|
2784 |
|
---|
2785 | </span>Users search a full-text retrieval system based on the text transcripts
|
---|
2786 |
|
---|
2787 | to retrieve relevant transcript sections and their associated audio recordings
|
---|
2788 |
|
---|
2789 | and photographs.<span style='mso-spacerun:yes'>Â </span>It is also possible to
|
---|
2790 |
|
---|
2791 | search for photos by matching text queries against text descriptions of the
|
---|
2792 |
|
---|
2793 | photos in the collection, where the located photos link back to their
|
---|
2794 |
|
---|
2795 | respective interview transcript and audio recordings.</span></p>
|
---|
2796 |
|
---|
2797 |
|
---|
2798 |
|
---|
2799 |
|
---|
2800 |
|
---|
2801 |
|
---|
2802 |
|
---|
2803 |
|
---|
2804 |
|
---|
2805 |
|
---|
2806 |
|
---|
2807 |
|
---|
2808 |
|
---|
2809 |
|
---|
2810 |
|
---|
2811 |
|
---|
2812 |
|
---|
2813 |
|
---|
2814 |
|
---|
2815 | <p class=MsoNormal style='margin-right:-.4pt'><b style='mso-bidi-font-weight:
|
---|
2816 |
|
---|
2817 | normal'><span lang=EN-GB>1997<o:p></o:p></span></b></p>
|
---|
2818 |
|
---|
2819 |
|
---|
2820 |
|
---|
2821 |
|
---|
2822 |
|
---|
2823 |
|
---|
2824 |
|
---|
2825 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/1</span></p>
|
---|
2826 |
|
---|
2827 |
|
---|
2828 |
|
---|
2829 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>A dynamic and
|
---|
2830 |
|
---|
2831 | flexible representation of social relationships in CSCW</span></p>
|
---|
2832 |
|
---|
2833 |
|
---|
2834 |
|
---|
2835 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Steve Jones,
|
---|
2836 |
|
---|
2837 | Steve Marsh</span></p>
|
---|
2838 |
|
---|
2839 |
|
---|
2840 |
|
---|
2841 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>CSCW system
|
---|
2842 |
|
---|
2843 | designers lack effective support in addressing the social issues and
|
---|
2844 |
|
---|
2845 | interpersonal relationships which are linked with the use of CSCW systems.<span
|
---|
2846 |
|
---|
2847 | style='mso-spacerun:yes'>Â </span>We present a formal description of trust to
|
---|
2848 |
|
---|
2849 | support CSCW system designers in considering the social aspects of group work,
|
---|
2850 |
|
---|
2851 | embedding those considerations in systems and analysing computer supported
|
---|
2852 |
|
---|
2853 | group processes.</span></p>
|
---|
2854 |
|
---|
2855 |
|
---|
2856 |
|
---|
2857 |
|
---|
2858 |
|
---|
2859 |
|
---|
2860 |
|
---|
2861 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We argue that
|
---|
2862 |
|
---|
2863 | trust is a critical aspect in group work, and describe what we consider to be
|
---|
2864 |
|
---|
2865 | the building blocks of trust.<span style='mso-spacerun:yes'>Â </span>We then
|
---|
2866 |
|
---|
2867 | present a formal notation for the building blocks, their use in reasoning about
|
---|
2868 |
|
---|
2869 | social interactions and how they are amended over time.</span></p>
|
---|
2870 |
|
---|
2871 |
|
---|
2872 |
|
---|
2873 |
|
---|
2874 |
|
---|
2875 |
|
---|
2876 |
|
---|
2877 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We then consider
|
---|
2878 |
|
---|
2879 | how the formalism may be used in practice, and present some insights from
|
---|
2880 |
|
---|
2881 | initial analysis of the behaviour of the formalism.<span
|
---|
2882 |
|
---|
2883 | style='mso-spacerun:yes'>Â </span>This is followed by a description of possible
|
---|
2884 |
|
---|
2885 | amendments and extensions to the formalism.<span style='mso-spacerun:yes'>Â
|
---|
2886 |
|
---|
2887 | </span>We conclude that it is possible to formalise a notion of trust and to
|
---|
2888 |
|
---|
2889 | model the formalisation by a computational mechanism.</span></p>
|
---|
2890 |
|
---|
2891 |
|
---|
2892 |
|
---|
2893 |
|
---|
2894 |
|
---|
2895 |
|
---|
2896 |
|
---|
2897 |
|
---|
2898 |
|
---|
2899 |
|
---|
2900 |
|
---|
2901 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/2</span></p>
|
---|
2902 |
|
---|
2903 |
|
---|
2904 |
|
---|
2905 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Design issues
|
---|
2906 |
|
---|
2907 | for World Wide Web navigation visualisation tools</span></p>
|
---|
2908 |
|
---|
2909 |
|
---|
2910 |
|
---|
2911 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Andy <span
|
---|
2912 |
|
---|
2913 | class=SpellE>Cockburn</span>, Steve Jones</span></p>
|
---|
2914 |
|
---|
2915 |
|
---|
2916 |
|
---|
2917 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The World Wide
|
---|
2918 |
|
---|
2919 | Web (WWW) is a successful hypermedia information space used by millions of
|
---|
2920 |
|
---|
2921 | people, yet it suffers from many deficiencies and problems in support for
|
---|
2922 |
|
---|
2923 | navigation around its vast information space.<span style='mso-spacerun:yes'>Â
|
---|
2924 |
|
---|
2925 | </span>In this paper we identify the origins of these navigation problems,
|
---|
2926 |
|
---|
2927 | namely WWW browser design, WWW page design, and WWW page description
|
---|
2928 |
|
---|
2929 | languages.<span style='mso-spacerun:yes'>Â </span>Regardless of their origins,
|
---|
2930 |
|
---|
2931 | these problems are eventually represented to the user at the browserâs user
|
---|
2932 |
|
---|
2933 | interface.<span style='mso-spacerun:yes'>Â </span>To help overcome these
|
---|
2934 |
|
---|
2935 | problems, many tools are being developed which allow users to visualise WWW
|
---|
2936 |
|
---|
2937 | subspaces.<span style='mso-spacerun:yes'>Â </span>We identify five key issues
|
---|
2938 |
|
---|
2939 | in the design and functionality of these visualisation systems: characteristics
|
---|
2940 |
|
---|
2941 | of the visual representation, the scope of the subspace representation, the
|
---|
2942 |
|
---|
2943 | mechanisms for generating the visualisation, the degree of browser
|
---|
2944 |
|
---|
2945 | independence, and the navigation support facilities.<span
|
---|
2946 |
|
---|
2947 | style='mso-spacerun:yes'>Â </span>We provide a critical review of the diverse
|
---|
2948 |
|
---|
2949 | range of WWW visualisation tools with respect to these issues.</span></p>
|
---|
2950 |
|
---|
2951 |
|
---|
2952 |
|
---|
2953 |
|
---|
2954 |
|
---|
2955 |
|
---|
2956 |
|
---|
2957 |
|
---|
2958 |
|
---|
2959 |
|
---|
2960 |
|
---|
2961 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/3</span></p>
|
---|
2962 |
|
---|
2963 |
|
---|
2964 |
|
---|
2965 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Stacked <span
|
---|
2966 |
|
---|
2967 | class=SpellE>generalization</span>:<span style='mso-spacerun:yes'>Â </span>when
|
---|
2968 |
|
---|
2969 | does it work?</span></p>
|
---|
2970 |
|
---|
2971 |
|
---|
2972 |
|
---|
2973 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Kai <span
|
---|
2974 |
|
---|
2975 | class=SpellE>Ming</span> Ting, Ian H. <span class=SpellE>Witten</span></span></p>
|
---|
2976 |
|
---|
2977 |
|
---|
2978 |
|
---|
2979 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Stacked <span
|
---|
2980 |
|
---|
2981 | class=SpellE>generalization</span> is a general method of using a high-level
|
---|
2982 |
|
---|
2983 | model to combine lower-level models to achieve greater predictive
|
---|
2984 |
|
---|
2985 | accuracy.<span style='mso-spacerun:yes'>Â </span>In this paper we address two
|
---|
2986 |
|
---|
2987 | crucial issues which have been considered to be a 'black art' in classification
|
---|
2988 |
|
---|
2989 | tasks ever since the introduction of stacked <span class=SpellE>generalization</span>
|
---|
2990 |
|
---|
2991 | in 1992 by <span class=SpellE>Wolpert</span>: the type of <span class=SpellE>generalizer</span>
|
---|
2992 |
|
---|
2993 | that is suitable to derive the higher-level model, and the kind of attributes
|
---|
2994 |
|
---|
2995 | that should be used as its input. </span></p>
|
---|
2996 |
|
---|
2997 |
|
---|
2998 |
|
---|
2999 |
|
---|
3000 |
|
---|
3001 |
|
---|
3002 |
|
---|
3003 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We demonstrate
|
---|
3004 |
|
---|
3005 | the effectiveness of stacked <span class=SpellE>generalization</span> for
|
---|
3006 |
|
---|
3007 | combining three different types of learning algorithms, and also for combining
|
---|
3008 |
|
---|
3009 | models of the same type derived from a single learning algorithm in a
|
---|
3010 |
|
---|
3011 | multiple-data-batches scenario.<span style='mso-spacerun:yes'>Â </span>We also
|
---|
3012 |
|
---|
3013 | compare the performance of stacked <span class=SpellE>generalization</span>
|
---|
3014 |
|
---|
3015 | with published results arcing and bagging.</span></p>
|
---|
3016 |
|
---|
3017 |
|
---|
3018 |
|
---|
3019 |
|
---|
3020 |
|
---|
3021 |
|
---|
3022 |
|
---|
3023 |
|
---|
3024 |
|
---|
3025 |
|
---|
3026 |
|
---|
3027 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/4</span></p>
|
---|
3028 |
|
---|
3029 |
|
---|
3030 |
|
---|
3031 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Browsing in
|
---|
3032 |
|
---|
3033 | digital libraries:<span style='mso-spacerun:yes'>Â </span>a phrase-based
|
---|
3034 |
|
---|
3035 | approach</span></p>
|
---|
3036 |
|
---|
3037 |
|
---|
3038 |
|
---|
3039 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Craig <span
|
---|
3040 |
|
---|
3041 | class=SpellE>Nevill</span>-Manning, Ian H. <span class=SpellE>Witten</span>,
|
---|
3042 |
|
---|
3043 | Gordon W. <span class=SpellE>Paynter</span></span></p>
|
---|
3044 |
|
---|
3045 |
|
---|
3046 |
|
---|
3047 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>A key question
|
---|
3048 |
|
---|
3049 | for digital libraries is this: how should one go about becoming familiar with a
|
---|
3050 |
|
---|
3051 | digital collection, as opposed to a physical one?<span
|
---|
3052 |
|
---|
3053 | style='mso-spacerun:yes'>Â </span>Digital collections generally present an
|
---|
3054 |
|
---|
3055 | appearance which is extremely opaque-a screen, typically a Web page, with no
|
---|
3056 |
|
---|
3057 | indication of what, or how much, lies beyond: whether a carefully-selected
|
---|
3058 |
|
---|
3059 | collection or a morass of worthless ephemera; whether half a dozen documents or
|
---|
3060 |
|
---|
3061 | many millions.<span style='mso-spacerun:yes'>Â </span>At least physical
|
---|
3062 |
|
---|
3063 | collections occupy physical space, present a physical appearance, and exhibit
|
---|
3064 |
|
---|
3065 | tangible physical <span class=SpellE>organization</span>.<span
|
---|
3066 |
|
---|
3067 | style='mso-spacerun:yes'>Â </span>When standing on the threshold of a large
|
---|
3068 |
|
---|
3069 | library one gains a sense of presence and permanence that reflects the care
|
---|
3070 |
|
---|
3071 | taken in building and maintaining the collection inside.<span
|
---|
3072 |
|
---|
3073 | style='mso-spacerun:yes'>Â </span>No-one could confuse it with a
|
---|
3074 |
|
---|
3075 | dung-heap!<span style='mso-spacerun:yes'>Â </span>Yet in the digital world the
|
---|
3076 |
|
---|
3077 | difference is not so palpable.</span></p>
|
---|
3078 |
|
---|
3079 |
|
---|
3080 |
|
---|
3081 |
|
---|
3082 |
|
---|
3083 |
|
---|
3084 |
|
---|
3085 |
|
---|
3086 |
|
---|
3087 |
|
---|
3088 |
|
---|
3089 |
|
---|
3090 |
|
---|
3091 |
|
---|
3092 |
|
---|
3093 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/5</span></p>
|
---|
3094 |
|
---|
3095 |
|
---|
3096 |
|
---|
3097 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>A graphical
|
---|
3098 |
|
---|
3099 | notation for the design of information visualisations</span></p>
|
---|
3100 |
|
---|
3101 |
|
---|
3102 |
|
---|
3103 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Matthew C. <span
|
---|
3104 |
|
---|
3105 | class=SpellE>Humphrey</span></span></p>
|
---|
3106 |
|
---|
3107 |
|
---|
3108 |
|
---|
3109 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Visualisations
|
---|
3110 |
|
---|
3111 | are coherent, graphical expressions of complex information that enhance peopleâs
|
---|
3112 |
|
---|
3113 | ability to communicate and reason about that information.<span
|
---|
3114 |
|
---|
3115 | style='mso-spacerun:yes'>Â </span>Yet despite the importance of visualisations
|
---|
3116 |
|
---|
3117 | in helping people to understand and solve a wide variety of problems, there is
|
---|
3118 |
|
---|
3119 | a dearth of formal tools and methods for discussing, describing and designing
|
---|
3120 |
|
---|
3121 | them.<span style='mso-spacerun:yes'>Â </span>Although simple visualisations,
|
---|
3122 |
|
---|
3123 | such as bar charts and <span class=SpellE>scatterplots</span>, are easily
|
---|
3124 |
|
---|
3125 | produced by modern interactive software, novel visualisations of multivariate, <span
|
---|
3126 |
|
---|
3127 | class=SpellE>multirelational</span> data must be expressed in a programming
|
---|
3128 |
|
---|
3129 | language.<span style='mso-spacerun:yes'>Â </span>The Relational Visualisation
|
---|
3130 |
|
---|
3131 | Notation is a new, graphical language for designing such highly expressive
|
---|
3132 |
|
---|
3133 | visualisations that does not use programming constructs.<span
|
---|
3134 |
|
---|
3135 | style='mso-spacerun:yes'>Â </span>Instead, the notation is based on relational
|
---|
3136 |
|
---|
3137 | algebra, which is widely used in database query languages, and it is supported
|
---|
3138 |
|
---|
3139 | by a suite of direct manipulation tools.<span style='mso-spacerun:yes'>Â
|
---|
3140 |
|
---|
3141 | </span>This article presents the notation and examines the designs of some
|
---|
3142 |
|
---|
3143 | interesting visualisations.</span></p>
|
---|
3144 |
|
---|
3145 |
|
---|
3146 |
|
---|
3147 |
|
---|
3148 |
|
---|
3149 |
|
---|
3150 |
|
---|
3151 |
|
---|
3152 |
|
---|
3153 |
|
---|
3154 |
|
---|
3155 |
|
---|
3156 |
|
---|
3157 |
|
---|
3158 |
|
---|
3159 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/6</span></p>
|
---|
3160 |
|
---|
3161 |
|
---|
3162 |
|
---|
3163 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Applications of
|
---|
3164 |
|
---|
3165 | machine learning in information retrieval</span></p>
|
---|
3166 |
|
---|
3167 |
|
---|
3168 |
|
---|
3169 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Sally Jo
|
---|
3170 |
|
---|
3171 | Cunningham, James <span class=SpellE>Littin</span>, Ian H. <span class=SpellE>Witten</span></span></p>
|
---|
3172 |
|
---|
3173 |
|
---|
3174 |
|
---|
3175 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Information
|
---|
3176 |
|
---|
3177 | retrieval systems provide access to collections of thousands, or millions, of
|
---|
3178 |
|
---|
3179 | documents, from which, by providing an appropriate description, users can
|
---|
3180 |
|
---|
3181 | recover any one.<span style='mso-spacerun:yes'>Â </span>Typically, users <span
|
---|
3182 |
|
---|
3183 | class=SpellE>iteratively</span> refine the descriptions they provide to satisfy
|
---|
3184 |
|
---|
3185 | their needs, and retrieval systems can <span class=SpellE>utilize</span> user
|
---|
3186 |
|
---|
3187 | feedback on selected documents to indicate the accuracy of the description at
|
---|
3188 |
|
---|
3189 | any stage.<span style='mso-spacerun:yes'>Â </span>The style of description
|
---|
3190 |
|
---|
3191 | required from the user, and the way it is employed to search the document
|
---|
3192 |
|
---|
3193 | database, are consequences of the indexing method used for the collection.<span
|
---|
3194 |
|
---|
3195 | style='mso-spacerun:yes'>Â </span>The index may take different forms, from
|
---|
3196 |
|
---|
3197 | storing keywords with links to individual documents, to clustering documents
|
---|
3198 |
|
---|
3199 | under related topics.</span></p>
|
---|
3200 |
|
---|
3201 |
|
---|
3202 |
|
---|
3203 |
|
---|
3204 |
|
---|
3205 |
|
---|
3206 |
|
---|
3207 |
|
---|
3208 |
|
---|
3209 |
|
---|
3210 |
|
---|
3211 |
|
---|
3212 |
|
---|
3213 |
|
---|
3214 |
|
---|
3215 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/7</span></p>
|
---|
3216 |
|
---|
3217 |
|
---|
3218 |
|
---|
3219 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Computer
|
---|
3220 |
|
---|
3221 | concepts without computers:<span style='mso-spacerun:yes'>Â </span>a first
|
---|
3222 |
|
---|
3223 | course in computer science</span></p>
|
---|
3224 |
|
---|
3225 |
|
---|
3226 |
|
---|
3227 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Geoffrey Holmes,
|
---|
3228 |
|
---|
3229 | Tony C. Smith, William J. Rogers</span></p>
|
---|
3230 |
|
---|
3231 |
|
---|
3232 |
|
---|
3233 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>While some
|
---|
3234 |
|
---|
3235 | institutions seek to make CS1 curricula more enjoyable by incorporating
|
---|
3236 |
|
---|
3237 | specialised educational software [1] or by setting more enjoyable programming
|
---|
3238 |
|
---|
3239 | assignments [2], we have joined the growing number of Computer Science
|
---|
3240 |
|
---|
3241 | departments that seek to improve the quality of the CS1 experience by focusing
|
---|
3242 |
|
---|
3243 | student attention away from the computer monitor [3,4].<span
|
---|
3244 |
|
---|
3245 | style='mso-spacerun:yes'>Â </span>Sophisticated computing concepts usually
|
---|
3246 |
|
---|
3247 | reserved for senior level courses are presented in a &lt;I&gt;popular
|
---|
3248 |
|
---|
3249 | science&lt;/I&gt; manner, and given equal time alongside the essential
|
---|
3250 |
|
---|
3251 | introductory programming material.<span style='mso-spacerun:yes'>Â </span>By
|
---|
3252 |
|
---|
3253 | exposing students to a broad range of specific computational problems we
|
---|
3254 |
|
---|
3255 | endeavour to make the introductory course more interesting and enjoyable, and
|
---|
3256 |
|
---|
3257 | instil in students a sense of vision for areas they might specialise in as
|
---|
3258 |
|
---|
3259 | computing majors.</span></p>
|
---|
3260 |
|
---|
3261 |
|
---|
3262 |
|
---|
3263 |
|
---|
3264 |
|
---|
3265 |
|
---|
3266 |
|
---|
3267 |
|
---|
3268 |
|
---|
3269 |
|
---|
3270 |
|
---|
3271 |
|
---|
3272 |
|
---|
3273 |
|
---|
3274 |
|
---|
3275 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/8</span></p>
|
---|
3276 |
|
---|
3277 |
|
---|
3278 |
|
---|
3279 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>A sight-singing
|
---|
3280 |
|
---|
3281 | tutor</span></p>
|
---|
3282 |
|
---|
3283 |
|
---|
3284 |
|
---|
3285 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Lloyd A. Smith, <span
|
---|
3286 |
|
---|
3287 | class=SpellE>Rodger</span> J. <span class=SpellE>McNab</span></span></p>
|
---|
3288 |
|
---|
3289 |
|
---|
3290 |
|
---|
3291 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
3292 |
|
---|
3293 | describes a computer program designed to aid its users in learning to
|
---|
3294 |
|
---|
3295 | sight-sing.<span style='mso-spacerun:yes'>Â </span>Sight-singing-the ability to
|
---|
3296 |
|
---|
3297 | sing music from a score without prior study-is an important skill for musicians
|
---|
3298 |
|
---|
3299 | and holds a central place in most university music curricula.<span
|
---|
3300 |
|
---|
3301 | style='mso-spacerun:yes'>Â </span>Its importance to vocalists is obvious; it is
|
---|
3302 |
|
---|
3303 | also an important skill for instrumentalists and conductors because it develops
|
---|
3304 |
|
---|
3305 | the aural imagination necessary to judge how the music should sound, when
|
---|
3306 |
|
---|
3307 | played (<span class=SpellE>Benward</span> and Carr 1991).<span
|
---|
3308 |
|
---|
3309 | style='mso-spacerun:yes'>Â </span>Furthermore, it is an important skill for
|
---|
3310 |
|
---|
3311 | amateur musicians, who can save a great deal of rehearsal time through an
|
---|
3312 |
|
---|
3313 | ability to sing music at sight.</span></p>
|
---|
3314 |
|
---|
3315 |
|
---|
3316 |
|
---|
3317 |
|
---|
3318 |
|
---|
3319 |
|
---|
3320 |
|
---|
3321 |
|
---|
3322 |
|
---|
3323 |
|
---|
3324 |
|
---|
3325 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/9</span></p>
|
---|
3326 |
|
---|
3327 |
|
---|
3328 |
|
---|
3329 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Stacking bagged
|
---|
3330 |
|
---|
3331 | and <span class=SpellE>dagged</span> models</span></p>
|
---|
3332 |
|
---|
3333 |
|
---|
3334 |
|
---|
3335 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Kai <span
|
---|
3336 |
|
---|
3337 | class=SpellE>Ming</span> Ting, I.H. <span class=SpellE>Witten</span></span></p>
|
---|
3338 |
|
---|
3339 |
|
---|
3340 |
|
---|
3341 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>In this paper,
|
---|
3342 |
|
---|
3343 | we investigate the method of <i style='mso-bidi-font-style:normal'>stacked <span
|
---|
3344 |
|
---|
3345 | class=SpellE>generalization</span></i> in combining models derived from
|
---|
3346 |
|
---|
3347 | different subsets of a training <span class=SpellE>dataset</span> by a single
|
---|
3348 |
|
---|
3349 | learning algorithm, as well as different algorithms.<span
|
---|
3350 |
|
---|
3351 | style='mso-spacerun:yes'>Â </span>The simplest way to combine predictions from
|
---|
3352 |
|
---|
3353 | competing models is majority vote, and the effect of the sampling regime used
|
---|
3354 |
|
---|
3355 | to generate training subsets has already been studied in this context-when
|
---|
3356 |
|
---|
3357 | bootstrap samples are used the method is called <i style='mso-bidi-font-style:
|
---|
3358 |
|
---|
3359 | normal'>bagging</i>, and for disjoint samples we call it <span class=SpellE><i
|
---|
3360 |
|
---|
3361 | style='mso-bidi-font-style:normal'>dagging</i></span>.<span
|
---|
3362 |
|
---|
3363 | style='mso-spacerun:yes'>Â </span>This paper extends these studies to stacked <span
|
---|
3364 |
|
---|
3365 | class=SpellE>generalization</span>, where a learning algorithm is employed to combine
|
---|
3366 |
|
---|
3367 | the models.<span style='mso-spacerun:yes'>Â </span>This yields new methods
|
---|
3368 |
|
---|
3369 | dubbed <i style='mso-bidi-font-style:normal'>bag-stacking</i> and <span
|
---|
3370 |
|
---|
3371 | class=SpellE><i style='mso-bidi-font-style:normal'>dag</i></span><i
|
---|
3372 |
|
---|
3373 | style='mso-bidi-font-style:normal'>-stacking</i>.</span></p>
|
---|
3374 |
|
---|
3375 |
|
---|
3376 |
|
---|
3377 |
|
---|
3378 |
|
---|
3379 |
|
---|
3380 |
|
---|
3381 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We demonstrate
|
---|
3382 |
|
---|
3383 | that bag-stacking and <span class=SpellE>dag</span>-stacking can be effective
|
---|
3384 |
|
---|
3385 | for classification tasks even when the training samples cover just a small
|
---|
3386 |
|
---|
3387 | fraction of the full <span class=SpellE>dataset</span>.<span
|
---|
3388 |
|
---|
3389 | style='mso-spacerun:yes'>Â </span>In contrast to earlier bagging results, we
|
---|
3390 |
|
---|
3391 | show that bagging and bag-stacking work for stable as well as unstable learning
|
---|
3392 |
|
---|
3393 | algorithms, as do <span class=SpellE>dagging</span> and <span class=SpellE>dag</span>-stacking.<span
|
---|
3394 |
|
---|
3395 | style='mso-spacerun:yes'>Â </span>We find that bag-stacking (<span
|
---|
3396 |
|
---|
3397 | class=SpellE>dag</span>-stacking) almost always has higher predictive accuracy
|
---|
3398 |
|
---|
3399 | than bagging (<span class=SpellE>dagging</span>), and we also show that
|
---|
3400 |
|
---|
3401 | bag-stacking models derived using two different algorithms is more effective
|
---|
3402 |
|
---|
3403 | than bagging.</span></p>
|
---|
3404 |
|
---|
3405 |
|
---|
3406 |
|
---|
3407 |
|
---|
3408 |
|
---|
3409 |
|
---|
3410 |
|
---|
3411 |
|
---|
3412 |
|
---|
3413 |
|
---|
3414 |
|
---|
3415 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/10</span></p>
|
---|
3416 |
|
---|
3417 |
|
---|
3418 |
|
---|
3419 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Extracting text
|
---|
3420 |
|
---|
3421 | from Postscript</span></p>
|
---|
3422 |
|
---|
3423 |
|
---|
3424 |
|
---|
3425 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Craig <span
|
---|
3426 |
|
---|
3427 | class=SpellE>Nevill</span>-Manning, Todd Reed, Ian H. <span class=SpellE>Witten</span></span></p>
|
---|
3428 |
|
---|
3429 |
|
---|
3430 |
|
---|
3431 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We show how to
|
---|
3432 |
|
---|
3433 | extract plain text from PostScript files. A textual scan is inadequate because
|
---|
3434 |
|
---|
3435 | PostScript interpreters can generate characters on the page that do not appear
|
---|
3436 |
|
---|
3437 | in the source file. Furthermore, word and line breaks are implicit in the
|
---|
3438 |
|
---|
3439 | graphical rendition, and must be inferred from the positioning of word
|
---|
3440 |
|
---|
3441 | fragments. We present a robust technique for extracting text and <span
|
---|
3442 |
|
---|
3443 | class=SpellE>recognizing</span> words and paragraphs. The method uses a
|
---|
3444 |
|
---|
3445 | standard PostScript interpreter but redefines several PostScript operators, and
|
---|
3446 |
|
---|
3447 | simple heuristics are employed to locate word and line breaks. The scheme has
|
---|
3448 |
|
---|
3449 | been used to create a full-text index, and plain-text versions, of 40,000
|
---|
3450 |
|
---|
3451 | technical reports (34 <span class=SpellE>Gbyte</span> of PostScript). Other
|
---|
3452 |
|
---|
3453 | text-extraction systems are reviewed: none offer the same combination of
|
---|
3454 |
|
---|
3455 | robustness and simplicity.</span></p>
|
---|
3456 |
|
---|
3457 |
|
---|
3458 |
|
---|
3459 |
|
---|
3460 |
|
---|
3461 |
|
---|
3462 |
|
---|
3463 |
|
---|
3464 |
|
---|
3465 |
|
---|
3466 |
|
---|
3467 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/11</span></p>
|
---|
3468 |
|
---|
3469 |
|
---|
3470 |
|
---|
3471 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Gathering and
|
---|
3472 |
|
---|
3473 | indexing rich fragments of the World Wide Web</span></p>
|
---|
3474 |
|
---|
3475 |
|
---|
3476 |
|
---|
3477 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Geoffrey Holmes,
|
---|
3478 |
|
---|
3479 | William J Rogers</span></p>
|
---|
3480 |
|
---|
3481 |
|
---|
3482 |
|
---|
3483 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>While the World
|
---|
3484 |
|
---|
3485 | Wide Web (WWW) is an attractive option as a resource for teaching and research
|
---|
3486 |
|
---|
3487 | it does have some undesirable features. The cost of allowing students unlimited
|
---|
3488 |
|
---|
3489 | access can be high-both in money and time; students may become addicted to
|
---|
3490 |
|
---|
3491 | 'surfing' the web-exploring purely for entertainment-and jeopardise their
|
---|
3492 |
|
---|
3493 | studies. Students are likely to discover undesirable material because large
|
---|
3494 |
|
---|
3495 | scale search engines index sites regardless of their merit. Finally, the
|
---|
3496 |
|
---|
3497 | explosive growth of WWW usage means that servers and networks are often
|
---|
3498 |
|
---|
3499 | overloaded, to the extent that a student may gain a very negative view of the
|
---|
3500 |
|
---|
3501 | technology.</span></p>
|
---|
3502 |
|
---|
3503 |
|
---|
3504 |
|
---|
3505 |
|
---|
3506 |
|
---|
3507 |
|
---|
3508 |
|
---|
3509 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We have developed
|
---|
3510 |
|
---|
3511 | a piece of software which attempts to address these issues by capturing rich
|
---|
3512 |
|
---|
3513 | fragments of the WWW onto local storage media. It is possible to put a
|
---|
3514 |
|
---|
3515 | collection onto CD ROM, providing portability and inexpensive storage. This
|
---|
3516 |
|
---|
3517 | enables the presentation of the WWW to distance learning students, who do not
|
---|
3518 |
|
---|
3519 | have internet access. The software interfaces to standard, commonly available
|
---|
3520 |
|
---|
3521 | web browsers, acting as a proxy server to the files stored on the local media,
|
---|
3522 |
|
---|
3523 | and provides a search engine giving full text searching capability within the
|
---|
3524 |
|
---|
3525 | collection.</span></p>
|
---|
3526 |
|
---|
3527 |
|
---|
3528 |
|
---|
3529 |
|
---|
3530 |
|
---|
3531 |
|
---|
3532 |
|
---|
3533 |
|
---|
3534 |
|
---|
3535 |
|
---|
3536 |
|
---|
3537 |
|
---|
3538 |
|
---|
3539 |
|
---|
3540 |
|
---|
3541 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/12</span></p>
|
---|
3542 |
|
---|
3543 |
|
---|
3544 |
|
---|
3545 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Using model
|
---|
3546 |
|
---|
3547 | trees for classification</span></p>
|
---|
3548 |
|
---|
3549 |
|
---|
3550 |
|
---|
3551 | <p class=MsoNormal style='margin-right:-.4pt'><span class=SpellE><span
|
---|
3552 |
|
---|
3553 | lang=EN-GB>Eibe</span></span><span lang=EN-GB> Frank, Yong Wang, Stuart <span
|
---|
3554 |
|
---|
3555 | class=SpellE>Inglis</span>, Geoffrey Holmes, Ian H. <span class=SpellE>Witten</span></span></p>
|
---|
3556 |
|
---|
3557 |
|
---|
3558 |
|
---|
3559 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Model trees,
|
---|
3560 |
|
---|
3561 | which are a type of decision tree with linear regression functions at the
|
---|
3562 |
|
---|
3563 | leaves, form the basis of a recent successful technique for predicting
|
---|
3564 |
|
---|
3565 | continuous numeric values.<span style='mso-spacerun:yes'>Â </span>They can be
|
---|
3566 |
|
---|
3567 | applied to classification problems by employing a standard method of
|
---|
3568 |
|
---|
3569 | transforming a classification problem into a problem of function
|
---|
3570 |
|
---|
3571 | approximation.<span style='mso-spacerun:yes'>Â </span>Surprisingly, using this
|
---|
3572 |
|
---|
3573 | simple transformation the model tree <span class=SpellE>inducer</span> M5',
|
---|
3574 |
|
---|
3575 | based on <span class=SpellE>Quinlan's</span> M5, generates more accurate
|
---|
3576 |
|
---|
3577 | classifiers than the state-of-the-art decision tree learner C5.0, particularly
|
---|
3578 |
|
---|
3579 | when most of the attributes are numeric.</span></p>
|
---|
3580 |
|
---|
3581 |
|
---|
3582 |
|
---|
3583 |
|
---|
3584 |
|
---|
3585 |
|
---|
3586 |
|
---|
3587 |
|
---|
3588 |
|
---|
3589 |
|
---|
3590 |
|
---|
3591 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/13</span></p>
|
---|
3592 |
|
---|
3593 |
|
---|
3594 |
|
---|
3595 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Discovering inter-attribute
|
---|
3596 |
|
---|
3597 | relationships</span></p>
|
---|
3598 |
|
---|
3599 |
|
---|
3600 |
|
---|
3601 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Geoffrey Holmes</span></p>
|
---|
3602 |
|
---|
3603 |
|
---|
3604 |
|
---|
3605 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>It is important
|
---|
3606 |
|
---|
3607 | to discover relationships between attributes being used to predict a class
|
---|
3608 |
|
---|
3609 | attribute in supervised learning situations for two reasons.<span
|
---|
3610 |
|
---|
3611 | style='mso-spacerun:yes'>Â </span>First, any such relationship will be
|
---|
3612 |
|
---|
3613 | potentially interesting to the provider of a <span class=SpellE>dataset</span>
|
---|
3614 |
|
---|
3615 | in its own right.<span style='mso-spacerun:yes'>Â </span>Second, it would
|
---|
3616 |
|
---|
3617 | simplify a learning algorithm's search space, and the related irrelevant
|
---|
3618 |
|
---|
3619 | feature and subset selection problem, if the relationships were removed from <span
|
---|
3620 |
|
---|
3621 | class=SpellE>datasets</span> ahead of learning.<span style='mso-spacerun:yes'>Â
|
---|
3622 |
|
---|
3623 | </span>An algorithm to discover such relationships is presented in this
|
---|
3624 |
|
---|
3625 | paper.<span style='mso-spacerun:yes'>Â </span>The algorithm is described and a
|
---|
3626 |
|
---|
3627 | surprising number of inter-attribute relationships are discovered in <span
|
---|
3628 |
|
---|
3629 | class=SpellE>datasets</span> from the University of California at Irvine (UCI)
|
---|
3630 |
|
---|
3631 | repository.</span></p>
|
---|
3632 |
|
---|
3633 |
|
---|
3634 |
|
---|
3635 |
|
---|
3636 |
|
---|
3637 |
|
---|
3638 |
|
---|
3639 |
|
---|
3640 |
|
---|
3641 |
|
---|
3642 |
|
---|
3643 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/14</span></p>
|
---|
3644 |
|
---|
3645 |
|
---|
3646 |
|
---|
3647 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Learning from <span
|
---|
3648 |
|
---|
3649 | class=SpellE>batched</span> data:<span style='mso-spacerun:yes'>Â </span>model
|
---|
3650 |
|
---|
3651 | combination <span class=SpellE>vs</span> data combination</span></p>
|
---|
3652 |
|
---|
3653 |
|
---|
3654 |
|
---|
3655 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Kai <span
|
---|
3656 |
|
---|
3657 | class=SpellE>Ming</span> Ting, Boon <span class=SpellE>Toh</span> Low, Ian H. <span
|
---|
3658 |
|
---|
3659 | class=SpellE>Witten</span></span></p>
|
---|
3660 |
|
---|
3661 |
|
---|
3662 |
|
---|
3663 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>When presented
|
---|
3664 |
|
---|
3665 | with multiple batches of data, one can either combine them into a single batch
|
---|
3666 |
|
---|
3667 | before applying a machine learning procedure or learn from each batch
|
---|
3668 |
|
---|
3669 | independently and combine the resulting models.<span style='mso-spacerun:yes'>Â
|
---|
3670 |
|
---|
3671 | </span>The former procedure, data combination, is straightforward; this paper
|
---|
3672 |
|
---|
3673 | investigates the latter, model combination.<span style='mso-spacerun:yes'>Â
|
---|
3674 |
|
---|
3675 | </span>Given an appropriate combination method, one might expect model
|
---|
3676 |
|
---|
3677 | combination to prove superior when the data in each batch was obtained under
|
---|
3678 |
|
---|
3679 | somewhat different conditions or when different learning algorithms were used
|
---|
3680 |
|
---|
3681 | on the batches.<span style='mso-spacerun:yes'>Â </span>Empirical results show
|
---|
3682 |
|
---|
3683 | that model combination often outperforms data combination even when the batches
|
---|
3684 |
|
---|
3685 | are drawn randomly from a single source of data and the same learning method is
|
---|
3686 |
|
---|
3687 | used on each.<span style='mso-spacerun:yes'>Â </span>Moreover, this is not just
|
---|
3688 |
|
---|
3689 | an <span class=SpellE>artifact</span> of one particular method of combining
|
---|
3690 |
|
---|
3691 | models: it occurs with several different combination methods.</span></p>
|
---|
3692 |
|
---|
3693 |
|
---|
3694 |
|
---|
3695 |
|
---|
3696 |
|
---|
3697 |
|
---|
3698 |
|
---|
3699 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We relate this
|
---|
3700 |
|
---|
3701 | phenomenon to the learning curve of the classifiers being used.<span
|
---|
3702 |
|
---|
3703 | style='mso-spacerun:yes'>Â </span>Early in the learning process when the
|
---|
3704 |
|
---|
3705 | learning curve is steep there is much to gain from data combination, but later
|
---|
3706 |
|
---|
3707 | when it becomes shallow there is less to gain and model combination achieves a
|
---|
3708 |
|
---|
3709 | greater reduction in variance and hence a lower error rate.</span></p>
|
---|
3710 |
|
---|
3711 |
|
---|
3712 |
|
---|
3713 |
|
---|
3714 |
|
---|
3715 |
|
---|
3716 |
|
---|
3717 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The practical
|
---|
3718 |
|
---|
3719 | implication of these results is that one should consider using model
|
---|
3720 |
|
---|
3721 | combination rather than data combination, especially when multiple batches of
|
---|
3722 |
|
---|
3723 | data for the same task are readily available.<span style='mso-spacerun:yes'>Â
|
---|
3724 |
|
---|
3725 | </span>It is often superior even when the batches are drawn randomly from a
|
---|
3726 |
|
---|
3727 | single sample, and we expect its advantage to increase if genuine statistical
|
---|
3728 |
|
---|
3729 | differences between the batches exist.</span></p>
|
---|
3730 |
|
---|
3731 |
|
---|
3732 |
|
---|
3733 |
|
---|
3734 |
|
---|
3735 |
|
---|
3736 |
|
---|
3737 |
|
---|
3738 |
|
---|
3739 |
|
---|
3740 |
|
---|
3741 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/15</span></p>
|
---|
3742 |
|
---|
3743 |
|
---|
3744 |
|
---|
3745 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Information
|
---|
3746 |
|
---|
3747 | seeking retrieval, reading and storing behaviour of library users</span></p>
|
---|
3748 |
|
---|
3749 |
|
---|
3750 |
|
---|
3751 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Turner K.</span></p>
|
---|
3752 |
|
---|
3753 |
|
---|
3754 |
|
---|
3755 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>In the interest
|
---|
3756 |
|
---|
3757 | of digital libraries, it is advisable that designers be aware of the potential
|
---|
3758 |
|
---|
3759 | behaviour of the users of such a system.<span style='mso-spacerun:yes'>Â
|
---|
3760 |
|
---|
3761 | </span>There are two distinct parts under investigation, the interaction
|
---|
3762 |
|
---|
3763 | between traditional libraries involving the seeking and retrieval of relevant
|
---|
3764 |
|
---|
3765 | material, and the reading and storage behaviours ensuing. Through this
|
---|
3766 |
|
---|
3767 | analysis, the findings could be incorporated into digital library facilities.
|
---|
3768 |
|
---|
3769 | There has been copious amounts of research on information seeking leading to
|
---|
3770 |
|
---|
3771 | the development of behavioural models to describe the process. Often research
|
---|
3772 |
|
---|
3773 | on the information seeking practices of individuals is based on the task and
|
---|
3774 |
|
---|
3775 | field of study. The information seeking model, presented by Ellis et al.
|
---|
3776 |
|
---|
3777 | (1993), characterises the format of this study where it is used to compare
|
---|
3778 |
|
---|
3779 | various research on the information seeking practices of groups of people (from
|
---|
3780 |
|
---|
3781 | academics to professionals). It is found that, although researchers do make use
|
---|
3782 |
|
---|
3783 | of library facilities, they tend to rely heavily on their own collections and
|
---|
3784 |
|
---|
3785 | primarily use the library as a source for previously identified information,
|
---|
3786 |
|
---|
3787 | browsing and <span class=SpellE>interloan</span>. It was found that there are
|
---|
3788 |
|
---|
3789 | significant differences in user behaviour between the groups analysed. When
|
---|
3790 |
|
---|
3791 | looking at the reading and storage of material it was hard to draw conclusions,
|
---|
3792 |
|
---|
3793 | due to the lack of substantial research and information on the topic. However,
|
---|
3794 |
|
---|
3795 | through the use of reading strategies, a general idea on how readers behave can
|
---|
3796 |
|
---|
3797 | be developed. Designers of digital libraries can benefit from the guidelines
|
---|
3798 |
|
---|
3799 | presented here to better understand their audience.</span></p>
|
---|
3800 |
|
---|
3801 |
|
---|
3802 |
|
---|
3803 |
|
---|
3804 |
|
---|
3805 |
|
---|
3806 |
|
---|
3807 |
|
---|
3808 |
|
---|
3809 |
|
---|
3810 |
|
---|
3811 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/16</span></p>
|
---|
3812 |
|
---|
3813 |
|
---|
3814 |
|
---|
3815 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Proceeding of
|
---|
3816 |
|
---|
3817 | the INTERACT97 Combined Workshop on CSCW in HCI-<span class=SpellE>Worldwide</span></span></p>
|
---|
3818 |
|
---|
3819 |
|
---|
3820 |
|
---|
3821 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Matthias <span
|
---|
3822 |
|
---|
3823 | class=SpellE>Rauterberg</span>, Lars <span class=SpellE>Oestreicher</span>,
|
---|
3824 |
|
---|
3825 | John <span class=SpellE>Grundy</span></span></p>
|
---|
3826 |
|
---|
3827 |
|
---|
3828 |
|
---|
3829 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This is the
|
---|
3830 |
|
---|
3831 | proceedings for the INTERACT97 combined workshop on âCSCW in HCI-<span
|
---|
3832 |
|
---|
3833 | class=SpellE>worldwide</span>â.<span style='mso-spacerun:yes'>Â </span>The
|
---|
3834 |
|
---|
3835 | position papers in this proceedings are those selected from topics relating to
|
---|
3836 |
|
---|
3837 | HCI community development <span class=SpellE>worldwide</span> and to CSCW
|
---|
3838 |
|
---|
3839 | issues.<span style='mso-spacerun:yes'>Â </span>Originally these were to be two
|
---|
3840 |
|
---|
3841 | separate INTERACT workshops, but were combined to ensure sufficient
|
---|
3842 |
|
---|
3843 | participation for a combined workshop to run.</span></p>
|
---|
3844 |
|
---|
3845 |
|
---|
3846 |
|
---|
3847 |
|
---|
3848 |
|
---|
3849 |
|
---|
3850 |
|
---|
3851 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The combined
|
---|
3852 |
|
---|
3853 | workshop has been split into two separate sessions to run in the morning of
|
---|
3854 |
|
---|
3855 | July 15<sup>th</sup>, Sydney, Australia.<span style='mso-spacerun:yes'>Â
|
---|
3856 |
|
---|
3857 | </span>One to discuss the issues relating to the position papers focusing on
|
---|
3858 |
|
---|
3859 | general CSCW systems, the other to the development of HCI communities in a <span
|
---|
3860 |
|
---|
3861 | class=SpellE>worldwide</span> context.<span style='mso-spacerun:yes'>Â
|
---|
3862 |
|
---|
3863 | </span>The CSCW session uses as a case study a proposed <span class=SpellE>groupware</span>
|
---|
3864 |
|
---|
3865 | tool for facilitating the development of an HCI database with a <span
|
---|
3866 |
|
---|
3867 | class=SpellE>worldwide</span> geographical distribution.<span
|
---|
3868 |
|
---|
3869 | style='mso-spacerun:yes'>Â </span>The HCI community session focuses on
|
---|
3870 |
|
---|
3871 | developing the content for such a database, in order for it to foster the
|
---|
3872 |
|
---|
3873 | continued development of HCI communities.<span style='mso-spacerun:yes'>Â
|
---|
3874 |
|
---|
3875 | </span>The afternoon session of the combined workshop involves a joint
|
---|
3876 |
|
---|
3877 | discussion of the case study <span class=SpellE>groupware</span> tool, in terms
|
---|
3878 |
|
---|
3879 | of its content and likely <span class=SpellE>groupware</span> facilities.</span></p>
|
---|
3880 |
|
---|
3881 |
|
---|
3882 |
|
---|
3883 |
|
---|
3884 |
|
---|
3885 |
|
---|
3886 |
|
---|
3887 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The position
|
---|
3888 |
|
---|
3889 | papers have been grouped into those focusing on HCI communities and hence
|
---|
3890 |
|
---|
3891 | content issues for a <span class=SpellE>groupware</span> database, and those focusing
|
---|
3892 |
|
---|
3893 | on CSCW and <span class=SpellE>groupware</span> issues, and hence likely <span
|
---|
3894 |
|
---|
3895 | class=SpellE>groupware</span> support in the proposed HCI
|
---|
3896 |
|
---|
3897 | database/collaboration tools.<span style='mso-spacerun:yes'>Â </span>We hope
|
---|
3898 |
|
---|
3899 | that you find the position papers in this proceedings offer a wide range of
|
---|
3900 |
|
---|
3901 | interesting reports of HCI community development <span class=SpellE>worldwide</span>,
|
---|
3902 |
|
---|
3903 | leading CSCW system research, and that a <span class=SpellE>groupware</span>
|
---|
3904 |
|
---|
3905 | tool supporting aspects of a <span class=SpellE>worldwide</span> HCI database
|
---|
3906 |
|
---|
3907 | can draw upon the varied work reported.</span></p>
|
---|
3908 |
|
---|
3909 |
|
---|
3910 |
|
---|
3911 |
|
---|
3912 |
|
---|
3913 |
|
---|
3914 |
|
---|
3915 |
|
---|
3916 |
|
---|
3917 |
|
---|
3918 |
|
---|
3919 |
|
---|
3920 |
|
---|
3921 |
|
---|
3922 |
|
---|
3923 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/17</span></p>
|
---|
3924 |
|
---|
3925 |
|
---|
3926 |
|
---|
3927 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Internationalising
|
---|
3928 |
|
---|
3929 | a spreadsheet for Pacific Basin languages</span></p>
|
---|
3930 |
|
---|
3931 |
|
---|
3932 |
|
---|
3933 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Robert <span
|
---|
3934 |
|
---|
3935 | class=SpellE>Barbour</span>, Alvin <span class=SpellE>Yeo</span></span></p>
|
---|
3936 |
|
---|
3937 |
|
---|
3938 |
|
---|
3939 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>As people trade
|
---|
3940 |
|
---|
3941 | and engage in commerce, an economically dominant culture tends to migrate
|
---|
3942 |
|
---|
3943 | language into other recently contacted cultures.<span
|
---|
3944 |
|
---|
3945 | style='mso-spacerun:yes'>Â </span>Information technology (IT) can accelerate <span
|
---|
3946 |
|
---|
3947 | class=SpellE>enculturation</span> and promote the expansion of western hegemony
|
---|
3948 |
|
---|
3949 | in IT.<span style='mso-spacerun:yes'>Â </span>Equally, IT can present a
|
---|
3950 |
|
---|
3951 | culturally appropriate interface to the user that promotes the preservation of
|
---|
3952 |
|
---|
3953 | culture and language with very little additional effort.<span
|
---|
3954 |
|
---|
3955 | style='mso-spacerun:yes'>Â </span>In this paper a spreadsheet is
|
---|
3956 |
|
---|
3957 | internationalised to accept languages from the Latin-1 character set such as
|
---|
3958 |
|
---|
3959 | English, <span class=SpellE>Maori</span> and <span class=SpellE>Bahasa</span> <span
|
---|
3960 |
|
---|
3961 | class=SpellE>Melayu</span> (Malaysiaâs national language).<span
|
---|
3962 |
|
---|
3963 | style='mso-spacerun:yes'>Â </span>A technique that allows a non-programmer to
|
---|
3964 |
|
---|
3965 | add a new language to the spreadsheet is described.<span
|
---|
3966 |
|
---|
3967 | style='mso-spacerun:yes'>Â </span>The technique could also be used to
|
---|
3968 |
|
---|
3969 | internationalise other software at the point of design by following the steps
|
---|
3970 |
|
---|
3971 | we outline.</span></p>
|
---|
3972 |
|
---|
3973 |
|
---|
3974 |
|
---|
3975 |
|
---|
3976 |
|
---|
3977 |
|
---|
3978 |
|
---|
3979 |
|
---|
3980 |
|
---|
3981 |
|
---|
3982 |
|
---|
3983 |
|
---|
3984 |
|
---|
3985 |
|
---|
3986 |
|
---|
3987 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/18</span></p>
|
---|
3988 |
|
---|
3989 |
|
---|
3990 |
|
---|
3991 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Localising a
|
---|
3992 |
|
---|
3993 | spreadsheet:<span style='mso-spacerun:yes'>Â </span>an <span class=SpellE>Iban</span>
|
---|
3994 |
|
---|
3995 | example</span></p>
|
---|
3996 |
|
---|
3997 |
|
---|
3998 |
|
---|
3999 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Alvin <span
|
---|
4000 |
|
---|
4001 | class=SpellE>Yeo</span>, Robert <span class=SpellE>Barbour</span></span></p>
|
---|
4002 |
|
---|
4003 |
|
---|
4004 |
|
---|
4005 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Presently, there
|
---|
4006 |
|
---|
4007 | is little localisation of software to smaller cultures if it is not
|
---|
4008 |
|
---|
4009 | economically viable.<span style='mso-spacerun:yes'>Â </span>We believe software
|
---|
4010 |
|
---|
4011 | should also be localised to the languages of small cultures in order to sustain
|
---|
4012 |
|
---|
4013 | and preserve these small cultures.<span style='mso-spacerun:yes'>Â </span>As an
|
---|
4014 |
|
---|
4015 | example, we localised a spreadsheet from English to <span class=SpellE>Iban</span>.<span
|
---|
4016 |
|
---|
4017 | style='mso-spacerun:yes'>Â </span>The process in which we carried out the
|
---|
4018 |
|
---|
4019 | localisation can be used as a framework for the localisation of software to
|
---|
4020 |
|
---|
4021 | languages of small ethnic minorities.<span style='mso-spacerun:yes'>Â
|
---|
4022 |
|
---|
4023 | </span>Some problems faced during the localisation process are also discussed.</span></p>
|
---|
4024 |
|
---|
4025 |
|
---|
4026 |
|
---|
4027 |
|
---|
4028 |
|
---|
4029 |
|
---|
4030 |
|
---|
4031 |
|
---|
4032 |
|
---|
4033 |
|
---|
4034 |
|
---|
4035 |
|
---|
4036 |
|
---|
4037 |
|
---|
4038 |
|
---|
4039 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/19</span></p>
|
---|
4040 |
|
---|
4041 |
|
---|
4042 |
|
---|
4043 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Strategies of
|
---|
4044 |
|
---|
4045 | internationalisation and localisation: a postmodernist/s perspective</span></p>
|
---|
4046 |
|
---|
4047 |
|
---|
4048 |
|
---|
4049 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Alvin <span
|
---|
4050 |
|
---|
4051 | class=SpellE>Yeo</span>, Robert <span class=SpellE>Barbour</span></span></p>
|
---|
4052 |
|
---|
4053 |
|
---|
4054 |
|
---|
4055 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Many software
|
---|
4056 |
|
---|
4057 | companies today are developing software not only for local consumption but for
|
---|
4058 |
|
---|
4059 | the rest of the world.<span style='mso-spacerun:yes'>Â </span>We introduce the
|
---|
4060 |
|
---|
4061 | concepts of internationalisation and localisation and discuss some techniques
|
---|
4062 |
|
---|
4063 | using these processes.<span style='mso-spacerun:yes'>Â </span>An examination of
|
---|
4064 |
|
---|
4065 | <span class=SpellE>postmodern</span> critique with respect to the software
|
---|
4066 |
|
---|
4067 | industry is also reported.<span style='mso-spacerun:yes'>Â </span>In addition,
|
---|
4068 |
|
---|
4069 | we also feature our proposed internationalisation technique that was inspired
|
---|
4070 |
|
---|
4071 | by taking into account the researches of <span class=SpellE>postmodern</span>
|
---|
4072 |
|
---|
4073 | philosophers and mathematicians.<span style='mso-spacerun:yes'>Â </span>As illustrated
|
---|
4074 |
|
---|
4075 | in our prototype, the technique empowers non-programmers to localise their own
|
---|
4076 |
|
---|
4077 | software.<span style='mso-spacerun:yes'>Â </span>Further development of the
|
---|
4078 |
|
---|
4079 | technique and its implications on user interfaces and the future of software
|
---|
4080 |
|
---|
4081 | internationalisation and localisation are discussed.</span></p>
|
---|
4082 |
|
---|
4083 |
|
---|
4084 |
|
---|
4085 |
|
---|
4086 |
|
---|
4087 |
|
---|
4088 |
|
---|
4089 |
|
---|
4090 |
|
---|
4091 |
|
---|
4092 |
|
---|
4093 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/20</span></p>
|
---|
4094 |
|
---|
4095 |
|
---|
4096 |
|
---|
4097 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Language use in
|
---|
4098 |
|
---|
4099 | software</span></p>
|
---|
4100 |
|
---|
4101 |
|
---|
4102 |
|
---|
4103 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Alvin <span
|
---|
4104 |
|
---|
4105 | class=SpellE>Yeo</span>, Robert <span class=SpellE>Barbour</span></span></p>
|
---|
4106 |
|
---|
4107 |
|
---|
4108 |
|
---|
4109 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Many of the
|
---|
4110 |
|
---|
4111 | popular software we use today are in English.<span style='mso-spacerun:yes'>Â
|
---|
4112 |
|
---|
4113 | </span>Very few software applications are available in minority languages.<span
|
---|
4114 |
|
---|
4115 | style='mso-spacerun:yes'>Â </span>Besides economic goals, we justify why
|
---|
4116 |
|
---|
4117 | software should be made available to smaller cultures.<span
|
---|
4118 |
|
---|
4119 | style='mso-spacerun:yes'>Â </span>Furthermore, there is evidence that people
|
---|
4120 |
|
---|
4121 | learn and progress faster in software in their mother tongue (<span
|
---|
4122 |
|
---|
4123 | class=SpellE>Griffiths</span> et at, 1994) (<span class=SpellE>Krock</span>,
|
---|
4124 |
|
---|
4125 | 1996).<span style='mso-spacerun:yes'>Â </span>We hypothesise that experienced
|
---|
4126 |
|
---|
4127 | users of English spreadsheet can easily migrate to a spreadsheet in their
|
---|
4128 |
|
---|
4129 | native tongue i.e. <span class=SpellE>Bahasa</span> <span class=SpellE>Melayu</span>
|
---|
4130 |
|
---|
4131 | (Malaysiaâs national language).<span style='mso-spacerun:yes'>Â
|
---|
4132 |
|
---|
4133 | </span>Observations made in the study suggest that the native speakers of <span
|
---|
4134 |
|
---|
4135 | class=SpellE>Bahasa</span> <span class=SpellE>Melayu</span> had difficulties
|
---|
4136 |
|
---|
4137 | with the <span class=SpellE>Bahasa</span> <span class=SpellE>Melayu</span>
|
---|
4138 |
|
---|
4139 | interface.<span style='mso-spacerun:yes'>Â </span>The subjectsâ main difficulty
|
---|
4140 |
|
---|
4141 | was their unfamiliarity with computing terminology in <span class=SpellE>Bahasa</span>
|
---|
4142 |
|
---|
4143 | <span class=SpellE>Melayu</span>.<span style='mso-spacerun:yes'>Â </span>We
|
---|
4144 |
|
---|
4145 | present possible strategies to increase the use of <span class=SpellE>Bahasa</span>
|
---|
4146 |
|
---|
4147 | <span class=SpellE>Melayu</span> in IT.<span style='mso-spacerun:yes'>Â
|
---|
4148 |
|
---|
4149 | </span>These strategies may also be used to promote the use of other minority
|
---|
4150 |
|
---|
4151 | languages in IT.</span></p>
|
---|
4152 |
|
---|
4153 |
|
---|
4154 |
|
---|
4155 |
|
---|
4156 |
|
---|
4157 |
|
---|
4158 |
|
---|
4159 |
|
---|
4160 |
|
---|
4161 |
|
---|
4162 |
|
---|
4163 |
|
---|
4164 |
|
---|
4165 |
|
---|
4166 |
|
---|
4167 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/21</span></p>
|
---|
4168 |
|
---|
4169 |
|
---|
4170 |
|
---|
4171 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Usability
|
---|
4172 |
|
---|
4173 | testing:<span style='mso-spacerun:yes'>Â </span>a Malaysian study</span></p>
|
---|
4174 |
|
---|
4175 |
|
---|
4176 |
|
---|
4177 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Alvin <span
|
---|
4178 |
|
---|
4179 | class=SpellE>Yeo</span>, Robert <span class=SpellE>Barbour</span>, Mark <span
|
---|
4180 |
|
---|
4181 | class=SpellE>Apperley</span></span></p>
|
---|
4182 |
|
---|
4183 |
|
---|
4184 |
|
---|
4185 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>An exploratory
|
---|
4186 |
|
---|
4187 | study of software assessment techniques is conducted in Malaysia.<span
|
---|
4188 |
|
---|
4189 | style='mso-spacerun:yes'>Â </span>Subjects in the study comprised staff members
|
---|
4190 |
|
---|
4191 | of a Malaysian university with a high Information Technology (IT) presence.<span
|
---|
4192 |
|
---|
4193 | style='mso-spacerun:yes'>Â </span>The subjects assessed a spreadsheet tool with
|
---|
4194 |
|
---|
4195 | a <span class=SpellE>Bahasa</span> <span class=SpellE>Melayu</span> (Malaysiaâs
|
---|
4196 |
|
---|
4197 | national language) interface.<span style='mso-spacerun:yes'>Â </span>Software
|
---|
4198 |
|
---|
4199 | evaluation techniques used include the think aloud method, interviews and the
|
---|
4200 |
|
---|
4201 | System Usability Scale.<span style='mso-spacerun:yes'>Â </span>The responses in
|
---|
4202 |
|
---|
4203 | the various techniques used are reported and initial results indicate
|
---|
4204 |
|
---|
4205 | idiosyncratic behaviour of Malaysian subjects.<span style='mso-spacerun:yes'>Â
|
---|
4206 |
|
---|
4207 | </span>The implications of the findings are also discussed.</span></p>
|
---|
4208 |
|
---|
4209 |
|
---|
4210 |
|
---|
4211 |
|
---|
4212 |
|
---|
4213 |
|
---|
4214 |
|
---|
4215 |
|
---|
4216 |
|
---|
4217 |
|
---|
4218 |
|
---|
4219 |
|
---|
4220 |
|
---|
4221 |
|
---|
4222 |
|
---|
4223 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/22</span></p>
|
---|
4224 |
|
---|
4225 |
|
---|
4226 |
|
---|
4227 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Inducing
|
---|
4228 |
|
---|
4229 | cost-sensitive trees via instance-weighting</span></p>
|
---|
4230 |
|
---|
4231 |
|
---|
4232 |
|
---|
4233 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Kai <span
|
---|
4234 |
|
---|
4235 | class=SpellE>Ming</span> Ting</span></p>
|
---|
4236 |
|
---|
4237 |
|
---|
4238 |
|
---|
4239 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>We introduce an
|
---|
4240 |
|
---|
4241 | instance-weighting method to induce cost-sensitive trees in this paper.<span
|
---|
4242 |
|
---|
4243 | style='mso-spacerun:yes'>Â </span>It is a <span class=SpellE>generalization</span>
|
---|
4244 |
|
---|
4245 | of the standard tree induction process where only the initial instance weights
|
---|
4246 |
|
---|
4247 | determine the type of tree (i.e., minimum error trees or minimum cost trees) to
|
---|
4248 |
|
---|
4249 | be induced.<span style='mso-spacerun:yes'>Â </span>We demonstrate that it can
|
---|
4250 |
|
---|
4251 | be easily adopted to an existing tree learning algorithm.</span></p>
|
---|
4252 |
|
---|
4253 |
|
---|
4254 |
|
---|
4255 |
|
---|
4256 |
|
---|
4257 |
|
---|
4258 |
|
---|
4259 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Previous
|
---|
4260 |
|
---|
4261 | research gave insufficient evidence to support the fact that the greedy
|
---|
4262 |
|
---|
4263 | divide-and-conquer algorithm can effectively induce a truly cost-sensitive tree
|
---|
4264 |
|
---|
4265 | directly from the training data.<span style='mso-spacerun:yes'>Â </span>We
|
---|
4266 |
|
---|
4267 | provide this empirical evidence in this paper.<span style='mso-spacerun:yes'>Â
|
---|
4268 |
|
---|
4269 | </span>The algorithm employing the instance-weighting method is found to be
|
---|
4270 |
|
---|
4271 | comparable to or better than both C4.5 and C5 in terms of total
|
---|
4272 |
|
---|
4273 | misclassification costs, tree size and the number of high cost errors.<span
|
---|
4274 |
|
---|
4275 | style='mso-spacerun:yes'>Â </span>The instance-weighting method is also simpler
|
---|
4276 |
|
---|
4277 | and more effective in implementation than a method based on altered priors.</span></p>
|
---|
4278 |
|
---|
4279 |
|
---|
4280 |
|
---|
4281 |
|
---|
4282 |
|
---|
4283 |
|
---|
4284 |
|
---|
4285 |
|
---|
4286 |
|
---|
4287 |
|
---|
4288 |
|
---|
4289 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/23</span></p>
|
---|
4290 |
|
---|
4291 |
|
---|
4292 |
|
---|
4293 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Fast convergence
|
---|
4294 |
|
---|
4295 | with a greedy tag-phrase dictionary</span></p>
|
---|
4296 |
|
---|
4297 |
|
---|
4298 |
|
---|
4299 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Ross <span
|
---|
4300 |
|
---|
4301 | class=SpellE>Peeters</span>, Tony C. Smith</span></p>
|
---|
4302 |
|
---|
4303 |
|
---|
4304 |
|
---|
4305 | <p class=MsoBodyText><span lang=EN-US>The best general-purpose compression
|
---|
4306 |
|
---|
4307 | schemes make their gains by estimating a probability distribution over all
|
---|
4308 |
|
---|
4309 | possible next symbols given the context established by some number of previous
|
---|
4310 |
|
---|
4311 | symbols.<span style='mso-spacerun:yes'>Â </span>Such context models typically
|
---|
4312 |
|
---|
4313 | obtain good compression results for plain text by taking advantage of
|
---|
4314 |
|
---|
4315 | regularities in character sequences.<span style='mso-spacerun:yes'>Â
|
---|
4316 |
|
---|
4317 | </span>Frequent words and syllables can be incorporated into the model quickly
|
---|
4318 |
|
---|
4319 | and thereafter used for reasonably accurate prediction.<span
|
---|
4320 |
|
---|
4321 | style='mso-spacerun:yes'>Â </span>However, the precise context in which
|
---|
4322 |
|
---|
4323 | frequent patterns emerge is often extremely varied, and each new word or phrase
|
---|
4324 |
|
---|
4325 | immediately introduces new contexts which can adversely affect the compression
|
---|
4326 |
|
---|
4327 | rate</span></p>
|
---|
4328 |
|
---|
4329 |
|
---|
4330 |
|
---|
4331 |
|
---|
4332 |
|
---|
4333 |
|
---|
4334 |
|
---|
4335 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>A great deal of
|
---|
4336 |
|
---|
4337 | the structural regularity in a natural language is given rather more by
|
---|
4338 |
|
---|
4339 | properties of its grammar than by the orthographic transcription of its
|
---|
4340 |
|
---|
4341 | phonology.<span style='mso-spacerun:yes'>Â </span>This implies that access to a
|
---|
4342 |
|
---|
4343 | grammatical abstraction might lead to good compression.<span
|
---|
4344 |
|
---|
4345 | style='mso-spacerun:yes'>Â </span>While grammatical models have been used
|
---|
4346 |
|
---|
4347 | successfully for compressing computer programs [4], grammar-based compression
|
---|
4348 |
|
---|
4349 | of plain text has received little attention, primarily because of the
|
---|
4350 |
|
---|
4351 | difficulties associated with constructing a suitable natural language
|
---|
4352 |
|
---|
4353 | grammar.<span style='mso-spacerun:yes'>Â </span>But even without a precise
|
---|
4354 |
|
---|
4355 | formulation of the syntax of a language, there is a linguistic abstraction
|
---|
4356 |
|
---|
4357 | which is easily accessed and which demonstrates a high degree of regularity
|
---|
4358 |
|
---|
4359 | which can be exploited for compression purposes-namely, lexical categories.</span></p>
|
---|
4360 |
|
---|
4361 |
|
---|
4362 |
|
---|
4363 |
|
---|
4364 |
|
---|
4365 |
|
---|
4366 |
|
---|
4367 |
|
---|
4368 |
|
---|
4369 |
|
---|
4370 |
|
---|
4371 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/24</span></p>
|
---|
4372 |
|
---|
4373 |
|
---|
4374 |
|
---|
4375 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Tag based models
|
---|
4376 |
|
---|
4377 | of English text</span></p>
|
---|
4378 |
|
---|
4379 |
|
---|
4380 |
|
---|
4381 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>W. J. <span
|
---|
4382 |
|
---|
4383 | class=SpellE>Teahan</span>, John G. <span class=SpellE>Cleary</span></span></p>
|
---|
4384 |
|
---|
4385 |
|
---|
4386 |
|
---|
4387 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The problem of
|
---|
4388 |
|
---|
4389 | compressing English text is important both because of the ubiquity of English
|
---|
4390 |
|
---|
4391 | as a target for compression and because of the light that compression can shed
|
---|
4392 |
|
---|
4393 | on the structure of English.<span style='mso-spacerun:yes'>Â </span>English
|
---|
4394 |
|
---|
4395 | text is examined in conjunction with additional information about the parts of
|
---|
4396 |
|
---|
4397 | speech of each word in the text (these are referred to as âtagsâ).<span
|
---|
4398 |
|
---|
4399 | style='mso-spacerun:yes'>Â </span>It is shown that the tags plus the text can
|
---|
4400 |
|
---|
4401 | be compressed more than the text alone.<span style='mso-spacerun:yes'>Â
|
---|
4402 |
|
---|
4403 | </span>Essentially the tags can be compressed for nothing or even a small net
|
---|
4404 |
|
---|
4405 | saving in size.<span style='mso-spacerun:yes'>Â </span>A comparison is made of
|
---|
4406 |
|
---|
4407 | a number of different ways of integrating compression of tags and text using an
|
---|
4408 |
|
---|
4409 | escape mechanism similar to PPM.<span style='mso-spacerun:yes'>Â </span>These
|
---|
4410 |
|
---|
4411 | are also compared with standard word based and character based compression
|
---|
4412 |
|
---|
4413 | programs.<span style='mso-spacerun:yes'>Â </span>The result is that the tag
|
---|
4414 |
|
---|
4415 | character and word based schemes always outperform the character based
|
---|
4416 |
|
---|
4417 | schemes.<span style='mso-spacerun:yes'>Â </span>Overall, the tag based schemes
|
---|
4418 |
|
---|
4419 | outperform the word based schemes.<span style='mso-spacerun:yes'>Â </span>We
|
---|
4420 |
|
---|
4421 | conclude by conjecturing that tags chosen for compression rather than
|
---|
4422 |
|
---|
4423 | linguistic purposes would perform even better.</span></p>
|
---|
4424 |
|
---|
4425 |
|
---|
4426 |
|
---|
4427 |
|
---|
4428 |
|
---|
4429 |
|
---|
4430 |
|
---|
4431 |
|
---|
4432 |
|
---|
4433 |
|
---|
4434 |
|
---|
4435 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/25</span></p>
|
---|
4436 |
|
---|
4437 |
|
---|
4438 |
|
---|
4439 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Musical image
|
---|
4440 |
|
---|
4441 | compression</span></p>
|
---|
4442 |
|
---|
4443 |
|
---|
4444 |
|
---|
4445 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>David
|
---|
4446 |
|
---|
4447 | Bainbridge, Stuart <span class=SpellE>Inglis</span></span></p>
|
---|
4448 |
|
---|
4449 |
|
---|
4450 |
|
---|
4451 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Optical music
|
---|
4452 |
|
---|
4453 | recognition aims to convert the vast repositories of sheet music in the world
|
---|
4454 |
|
---|
4455 | into an on-line digital format [Bai97].<span style='mso-spacerun:yes'>Â
|
---|
4456 |
|
---|
4457 | </span>In the near future it will be possible to assimilate music into digital
|
---|
4458 |
|
---|
4459 | libraries and users will be able to perform searches based on a sung melody in
|
---|
4460 |
|
---|
4461 | addition to typical text-based searching [MSW+96].<span
|
---|
4462 |
|
---|
4463 | style='mso-spacerun:yes'>Â </span>An important requirement for such a system is
|
---|
4464 |
|
---|
4465 | the ability to reproduce the original score as accurately as possible.<span
|
---|
4466 |
|
---|
4467 | style='mso-spacerun:yes'>Â </span>Due to the huge amount of sheet music
|
---|
4468 |
|
---|
4469 | available, the efficient storage of musical images is an important topic of
|
---|
4470 |
|
---|
4471 | study.</span></p>
|
---|
4472 |
|
---|
4473 |
|
---|
4474 |
|
---|
4475 |
|
---|
4476 |
|
---|
4477 |
|
---|
4478 |
|
---|
4479 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
4480 |
|
---|
4481 | investigates whether the âknowledgeâ extracted from the optical music
|
---|
4482 |
|
---|
4483 | recognition (OMR) process can be exploited to gain higher compression than the
|
---|
4484 |
|
---|
4485 | JBIG international standard for <span class=SpellE>bi</span>-level image
|
---|
4486 |
|
---|
4487 | compression.<span style='mso-spacerun:yes'>Â </span>We present a hybrid
|
---|
4488 |
|
---|
4489 | approach where the primitive shapes of music extracted by the optical music
|
---|
4490 |
|
---|
4491 | recognition process-note heads, note stems, staff lines and so forth-are fed
|
---|
4492 |
|
---|
4493 | into a graphical symbol based compression scheme originally designed for images
|
---|
4494 |
|
---|
4495 | containing mainly printed text.<span style='mso-spacerun:yes'>Â </span>Using
|
---|
4496 |
|
---|
4497 | this hybrid approach the average compression rate for a single page is improved
|
---|
4498 |
|
---|
4499 | by 3.5% over JBIG.<span style='mso-spacerun:yes'>Â </span>When multiple pages with
|
---|
4500 |
|
---|
4501 | similar typography are processed in sequence, the file size is decreased by
|
---|
4502 |
|
---|
4503 | 4-8%.</span></p>
|
---|
4504 |
|
---|
4505 |
|
---|
4506 |
|
---|
4507 |
|
---|
4508 |
|
---|
4509 |
|
---|
4510 |
|
---|
4511 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Section 2
|
---|
4512 |
|
---|
4513 | presents the relevant background to both optical music recognition and textual
|
---|
4514 |
|
---|
4515 | image compression.<span style='mso-spacerun:yes'>Â </span>Section 3 describes
|
---|
4516 |
|
---|
4517 | the experiments performed on 66 test images, outlining the combinations of
|
---|
4518 |
|
---|
4519 | parameters that were examined to give the best results.<span
|
---|
4520 |
|
---|
4521 | style='mso-spacerun:yes'>Â </span>The initial results and refinements are
|
---|
4522 |
|
---|
4523 | presented in Section 4, and we conclude in the last section by <span
|
---|
4524 |
|
---|
4525 | class=SpellE>summarizing</span> the findings of this work.</span></p>
|
---|
4526 |
|
---|
4527 |
|
---|
4528 |
|
---|
4529 |
|
---|
4530 |
|
---|
4531 |
|
---|
4532 |
|
---|
4533 |
|
---|
4534 |
|
---|
4535 |
|
---|
4536 |
|
---|
4537 |
|
---|
4538 |
|
---|
4539 |
|
---|
4540 |
|
---|
4541 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/26</span></p>
|
---|
4542 |
|
---|
4543 |
|
---|
4544 |
|
---|
4545 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Correcting English
|
---|
4546 |
|
---|
4547 | text using PPM models</span></p>
|
---|
4548 |
|
---|
4549 |
|
---|
4550 |
|
---|
4551 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>W. J. <span
|
---|
4552 |
|
---|
4553 | class=SpellE>Teahan</span>, S. <span class=SpellE>Inglis</span>, J. G. <span
|
---|
4554 |
|
---|
4555 | class=SpellE>Cleary</span>, G. Holmes</span></p>
|
---|
4556 |
|
---|
4557 |
|
---|
4558 |
|
---|
4559 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>An essential
|
---|
4560 |
|
---|
4561 | component of many applications in natural language processing is a language <span
|
---|
4562 |
|
---|
4563 | class=SpellE>modeler</span> able to correct errors in the text being
|
---|
4564 |
|
---|
4565 | processed.<span style='mso-spacerun:yes'>Â </span>For optical character recognition
|
---|
4566 |
|
---|
4567 | (OCR), poor scanning quality or extraneous pixels in the image may cause one or
|
---|
4568 |
|
---|
4569 | more characters to be mis-<span class=SpellE>recognized</span>; while for
|
---|
4570 |
|
---|
4571 | spelling correction, two characters may be transposed, or a character may be
|
---|
4572 |
|
---|
4573 | inadvertently inserted or missed out. </span></p>
|
---|
4574 |
|
---|
4575 |
|
---|
4576 |
|
---|
4577 |
|
---|
4578 |
|
---|
4579 |
|
---|
4580 |
|
---|
4581 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>This paper
|
---|
4582 |
|
---|
4583 | describes a method for correcting English text using a PPM model.<span
|
---|
4584 |
|
---|
4585 | style='mso-spacerun:yes'>Â </span>A method that segments words in English text
|
---|
4586 |
|
---|
4587 | is introduced and is shown to be a significant improvement over previously used
|
---|
4588 |
|
---|
4589 | methods.<span style='mso-spacerun:yes'>Â </span>A similar technique is also
|
---|
4590 |
|
---|
4591 | applied as a post-processing stage after pages have been <span class=SpellE>recognized</span>
|
---|
4592 |
|
---|
4593 | by a state-of-the-art commercial OCR system.<span style='mso-spacerun:yes'>Â
|
---|
4594 |
|
---|
4595 | </span>We show that the accuracy of the OCR system can be increased from 95.9%
|
---|
4596 |
|
---|
4597 | to 96.6%, a decrease of about 10 errors per page.</span></p>
|
---|
4598 |
|
---|
4599 |
|
---|
4600 |
|
---|
4601 |
|
---|
4602 |
|
---|
4603 |
|
---|
4604 |
|
---|
4605 |
|
---|
4606 |
|
---|
4607 |
|
---|
4608 |
|
---|
4609 |
|
---|
4610 |
|
---|
4611 |
|
---|
4612 |
|
---|
4613 |
|
---|
4614 |
|
---|
4615 |
|
---|
4616 |
|
---|
4617 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/27</span></p>
|
---|
4618 |
|
---|
4619 |
|
---|
4620 |
|
---|
4621 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Constraints on
|
---|
4622 |
|
---|
4623 | parallelism beyond 10 instructions per cycle</span></p>
|
---|
4624 |
|
---|
4625 |
|
---|
4626 |
|
---|
4627 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>John G. <span
|
---|
4628 |
|
---|
4629 | class=SpellE>Cleary</span>, Richard H. <span class=SpellE>Littin</span>, J. A.
|
---|
4630 |
|
---|
4631 | David <span class=SpellE>McWha</span>, Murray W. Pearson</span></p>
|
---|
4632 |
|
---|
4633 |
|
---|
4634 |
|
---|
4635 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The problem of
|
---|
4636 |
|
---|
4637 | extracting Instruction Level Parallelism at levels of 10 instructions per clock
|
---|
4638 |
|
---|
4639 | and higher is considered.<span style='mso-spacerun:yes'>Â </span>Two different
|
---|
4640 |
|
---|
4641 | architectures which use speculation on memory accesses to achieve this level of
|
---|
4642 |
|
---|
4643 | performance are reviewed.<span style='mso-spacerun:yes'>Â </span>It is pointed
|
---|
4644 |
|
---|
4645 | out that while this form of speculation gives high potential parallelism it is
|
---|
4646 |
|
---|
4647 | necessary to retain execution state so that incorrect speculation can be detected
|
---|
4648 |
|
---|
4649 | and subsequently squashed.<span style='mso-spacerun:yes'>Â </span>Simulation
|
---|
4650 |
|
---|
4651 | results show that the space to store such state is a critical resource in
|
---|
4652 |
|
---|
4653 | obtaining good speedup.<span style='mso-spacerun:yes'>Â </span>To make good use
|
---|
4654 |
|
---|
4655 | of the space it is essential that state be stored efficiently and that it be
|
---|
4656 |
|
---|
4657 | retired as soon as possible.<span style='mso-spacerun:yes'>Â </span>A number of
|
---|
4658 |
|
---|
4659 | techniques for extracting the best usage from the available state storage are
|
---|
4660 |
|
---|
4661 | introduced.</span></p>
|
---|
4662 |
|
---|
4663 |
|
---|
4664 |
|
---|
4665 |
|
---|
4666 |
|
---|
4667 |
|
---|
4668 |
|
---|
4669 |
|
---|
4670 |
|
---|
4671 |
|
---|
4672 |
|
---|
4673 |
|
---|
4674 |
|
---|
4675 |
|
---|
4676 |
|
---|
4677 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/28</span></p>
|
---|
4678 |
|
---|
4679 |
|
---|
4680 |
|
---|
4681 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Effects of
|
---|
4682 |
|
---|
4683 | re-ordered memory operations on parallelism</span></p>
|
---|
4684 |
|
---|
4685 |
|
---|
4686 |
|
---|
4687 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>Richard H. <span
|
---|
4688 |
|
---|
4689 | class=SpellE>Littin</span>, John G. <span class=SpellE>Cleary</span></span></p>
|
---|
4690 |
|
---|
4691 |
|
---|
4692 |
|
---|
4693 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>The performance
|
---|
4694 |
|
---|
4695 | effect of permitting different memory operations to be re-ordered is
|
---|
4696 |
|
---|
4697 | examined.<span style='mso-spacerun:yes'>Â </span>The available parallelism is
|
---|
4698 |
|
---|
4699 | computed using a machine code simulator.<span style='mso-spacerun:yes'>Â
|
---|
4700 |
|
---|
4701 | </span>A range of possible restrictions on the re-ordering of memory operations
|
---|
4702 |
|
---|
4703 | is considered: from the purely sequential case where no re-ordering is
|
---|
4704 |
|
---|
4705 | permitted; to the completely permissive one where memory operations may occur
|
---|
4706 |
|
---|
4707 | in any order so that the parallelism is restricted only by data
|
---|
4708 |
|
---|
4709 | dependencies.<span style='mso-spacerun:yes'>Â </span>A general conclusion is
|
---|
4710 |
|
---|
4711 | drawn that to reliably obtain parallelism beyond 10 instructions per clock will
|
---|
4712 |
|
---|
4713 | require an ability to re-order all memory instructions.<span
|
---|
4714 |
|
---|
4715 | style='mso-spacerun:yes'>Â </span>A brief description of a feasible
|
---|
4716 |
|
---|
4717 | architecture capable of this is given.</span></p>
|
---|
4718 |
|
---|
4719 |
|
---|
4720 |
|
---|
4721 |
|
---|
4722 |
|
---|
4723 |
|
---|
4724 |
|
---|
4725 |
|
---|
4726 |
|
---|
4727 |
|
---|
4728 |
|
---|
4729 |
|
---|
4730 |
|
---|
4731 |
|
---|
4732 |
|
---|
4733 | <p class=MsoNormal style='margin-right:-.4pt'><span lang=EN-GB>97/29</span></p>
|
---|
4734 |
|
---|
4735 |
|
---|
4736 |
|
---|
4737 | <p class=MsoNormal><span lang=EN-GB>OZCHIâ96 Industry Session:<span
|
---|
4738 |
|
---|
4739 | style='mso-spacerun:yes'>Â </span>Sixth Australian Conference on Human-Computer
|
---|
4740 |
|
---|
4741 | Interaction</span></p>
|
---|
4742 |
|
---|
4743 |
|
---|
4744 |
|
---|
4745 | <p class=MsoNormal><span lang=EN-GB>Edited by Chris Phillips, Janis <span
|
---|
4746 |
|
---|
4747 | class=SpellE>McKauge</span></span></p>
|
---|
4748 |
|
---|
4749 |
|
---|
4750 |
|
---|
4751 | <p class=MsoNormal><span lang=EN-GB>The idea for a specific industry session at
|
---|
4752 |
|
---|
4753 | OZCHI was first mooted at the 1995 conference in <span class=SpellE>Wollongong</span>,
|
---|
4754 |
|
---|
4755 | during questions following a session of short papers which happened
|
---|
4756 |
|
---|
4757 | (serendipitously) to be presented by people from industry.<span
|
---|
4758 |
|
---|
4759 | style='mso-spacerun:yes'>Â </span>An animated discussion took place, most of
|
---|
4760 |
|
---|
4761 | which was about how OZCHI could be made more relevant to people in industry, be
|
---|
4762 |
|
---|
4763 | it working as usability consultants, or working within organisations either as
|
---|
4764 |
|
---|
4765 | usability professionals or as âchampions of the causeâ.<span
|
---|
4766 |
|
---|
4767 | style='mso-spacerun:yes'>Â </span>The discussion raised more questions than
|
---|
4768 |
|
---|
4769 | answers, about the format of such as session, about the challenges of
|
---|
4770 |
|
---|
4771 | attracting industry participation, and about the best way of publishing the
|
---|
4772 |
|
---|
4773 | results.<span style='mso-spacerun:yes'>Â </span>Although no real solutions were
|
---|
4774 |
|
---|
4775 | arrived at, it was enough to place an industry session on the agenda for
|
---|
4776 |
|
---|
4777 | OZCHIâ96.</span></p>
|
---|
4778 |
|
---|
4779 |
|
---|
4780 |
|
---|
4781 |
|
---|
4782 |
|
---|
4783 |
|
---|
4784 |
|
---|
4785 |
|
---|
4786 |
|
---|
4787 |
|
---|
4788 |
|
---|
4789 | <p class=MsoNormal><span lang=EN-GB>97/30</span></p>
|
---|
4790 |
|
---|
4791 |
|
---|
4792 |
|
---|
4793 | <p class=MsoNormal><span lang=EN-GB>Adaptive models of English text</span></p>
|
---|
4794 |
|
---|
4795 |
|
---|
4796 |
|
---|
4797 | <p class=MsoNormal><span lang=EN-GB>W. J. <span class=SpellE>Teahan</span>,
|
---|
4798 |
|
---|
4799 | John G. <span class=SpellE>Cleary</span></span></p>
|
---|
4800 |
|
---|
4801 |
|
---|
4802 |
|
---|
4803 | <p class=MsoNormal><span lang=EN-GB>High quality models of English text with
|
---|
4804 |
|
---|
4805 | performance approaching that of humans is important for many applications
|
---|
4806 |
|
---|
4807 | including spelling correction, speech recognition, OCR, and encryption.<span
|
---|
4808 |
|
---|
4809 | style='mso-spacerun:yes'>Â </span>A number of different statistical models of
|
---|
4810 |
|
---|
4811 | English are compared with each other and with previous estimates from human
|
---|
4812 |
|
---|
4813 | subjects.<span style='mso-spacerun:yes'>Â </span>It is concluded that the best
|
---|
4814 |
|
---|
4815 | current models are word based with part of speech tags.<span
|
---|
4816 |
|
---|
4817 | style='mso-spacerun:yes'>Â </span>Given sufficient training text, they are able
|
---|
4818 |
|
---|
4819 | to attain performance comparable to humans.</span></p>
|
---|
4820 |
|
---|
4821 |
|
---|
4822 |
|
---|
4823 |
|
---|
4824 |
|
---|
4825 |
|
---|
4826 |
|
---|
4827 |
|
---|
4828 |
|
---|
4829 |
|
---|
4830 |
|
---|
4831 |
|
---|
4832 |
|
---|
4833 |
|
---|
4834 |
|
---|
4835 | <p class=MsoNormal><span lang=EN-GB>97/31</span></p>
|
---|
4836 |
|
---|
4837 |
|
---|
4838 |
|
---|
4839 | <p class=MsoNormal><span lang=EN-GB>A graphical user interface for Boolean
|
---|
4840 |
|
---|
4841 | query specification</span></p>
|
---|
4842 |
|
---|
4843 |
|
---|
4844 |
|
---|
4845 | <p class=MsoNormal><span lang=EN-GB>Steve Jones, <span class=SpellE>Shona</span>
|
---|
4846 |
|
---|
4847 | <span class=SpellE>McInnes</span></span></p>
|
---|
4848 |
|
---|
4849 |
|
---|
4850 |
|
---|
4851 | <p class=MsoNormal><span lang=EN-GB>On-line information repositories commonly
|
---|
4852 |
|
---|
4853 | provide keyword search facilities via textual query languages based on Boolean
|
---|
4854 |
|
---|
4855 | logic.<span style='mso-spacerun:yes'>Â </span>However, there is evidence to
|
---|
4856 |
|
---|
4857 | suggest that the syntactical demands of such languages can lead to user errors
|
---|
4858 |
|
---|
4859 | and adversely affect the time that it takes users to form queries.<span
|
---|
4860 |
|
---|
4861 | style='mso-spacerun:yes'>Â </span>Users also face difficulties because of the
|
---|
4862 |
|
---|
4863 | conflict in semantics between AND <span class=SpellE>and</span> OR when used in
|
---|
4864 |
|
---|
4865 | Boolean logic and English language.<span style='mso-spacerun:yes'>Â </span>We
|
---|
4866 |
|
---|
4867 | suggest that graphical query languages, in particular Venn-like diagrams, can
|
---|
4868 |
|
---|
4869 | alleviate the problems that users experience when forming Boolean expressions
|
---|
4870 |
|
---|
4871 | with textual languages.<span style='mso-spacerun:yes'>Â </span>We describe <span
|
---|
4872 |
|
---|
4873 | class=SpellE>Vquery</span>, a Venn-diagram based user interface to the New
|
---|
4874 |
|
---|
4875 | Zealand Digital Library (NZDL).<span style='mso-spacerun:yes'>Â </span>The
|
---|
4876 |
|
---|
4877 | design of <span class=SpellE>Vquery</span> has been partly motivated by
|
---|
4878 |
|
---|
4879 | analysis of NZDL usage.<span style='mso-spacerun:yes'>Â </span>We found that
|
---|
4880 |
|
---|
4881 | few queries contain more than three terms, use of the intersection operator
|
---|
4882 |
|
---|
4883 | dominates and that query refinement is common.<span style='mso-spacerun:yes'>Â
|
---|
4884 |
|
---|
4885 | </span>A study of the utility of Venn diagrams for query specification
|
---|
4886 |
|
---|
4887 | indicates that with little or no training users can interpret and form
|
---|
4888 |
|
---|
4889 | Venn-like diagrams which accurately correspond to Boolean expressions.<span
|
---|
4890 |
|
---|
4891 | style='mso-spacerun:yes'>Â </span>The utility of <span class=SpellE>Vquery</span>
|
---|
4892 |
|
---|
4893 | is considered and directions for future work are proposed.</span></p>
|
---|
4894 |
|
---|
4895 |
|
---|
4896 |
|
---|
4897 |
|
---|
4898 |
|
---|
4899 |
|
---|
4900 |
|
---|
4901 |
|
---|
4902 |
|
---|
4903 |
|
---|
4904 |
|
---|
4905 | </div>
|
---|
4906 |
|
---|
4907 |
|
---|
4908 |
|
---|
4909 |
|
---|
4910 |
|
---|
4911 |
|
---|
4912 |
|
---|
4913 |
|
---|
4914 |
|
---|
4915 | </Content>
|
---|
4916 | </Section>
|
---|
4917 | </Archive>
|
---|