source: other-projects/nightly-tasks/diffcol/trunk/gs3-model-collect/Word-PDF-Enhanced/archives/HASHeaa2.dir/doc.xml@ 30029

<?xml version="1.0" encoding="utf-8" standalone="no"?>
<!DOCTYPE Archive SYSTEM "http://greenstone.org/dtd/Archive/1.0/Archive.dtd">
<Archive>
<Section>
  <Description>
    <Metadata name="gsdldoctype">indexed_doc</Metadata>
    <Metadata name="Language">en</Metadata>
    <Metadata name="Encoding">windows_1252</Metadata>
    <Metadata name="Creator">dg5</Metadata>
    <Metadata name="Title">1997-00 Listing of Working Papers</Metadata>
    <Metadata name="URL">http://C:/Users/Anupama/GS307_13July2015/web/sites/localsite/collect/Word-PDF-Enhanced/tmp/1436775751/word01.html</Metadata>
    <Metadata name="UTF8URL">http://C:/Users/Anupama/GS307_13July2015/web/sites/localsite/collect/Word-PDF-Enhanced/tmp/1436775751/word01.html</Metadata>
    <Metadata name="gsdlsourcefilename">import\word01.doc</Metadata>
    <Metadata name="gsdlconvertedfilename">tmp\1436775751\word01.html</Metadata>
    <Metadata name="OrigSource">word01.html</Metadata>
    <Metadata name="Source">word01.doc</Metadata>
    <Metadata name="SourceFile">word01.doc</Metadata>
    <Metadata name="Plugin">WordPlugin</Metadata>
    <Metadata name="FileSize">110080</Metadata>
    <Metadata name="FilenameRoot">word01</Metadata>
    <Metadata name="FileFormat">Word</Metadata>
    <Metadata name="srcicon">_icondoc_</Metadata>
    <Metadata name="srclink_file">doc.doc</Metadata>
    <Metadata name="srclinkFile">doc.doc</Metadata>
    <Metadata name="Identifier">HASHeaa2992e081949673150f3</Metadata>
    <Metadata name="lastmodified">1436763858</Metadata>
    <Metadata name="lastmodifieddate">20150713</Metadata>
    <Metadata name="oailastmodified">1436775752</Metadata>
    <Metadata name="oailastmodifieddate">20150713</Metadata>
    <Metadata name="assocfilepath">HASHeaa2.dir</Metadata>
    <Metadata name="gsdlassocfile">doc.doc:application/msword:</Metadata>
  </Description>
  <Content>



&lt;div class=WordSection1&gt;



&lt;p class=MsoTitle&gt;&lt;span lang=EN-US&gt;1997-00 Listing of Working Papers &lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/1&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Using

compression to identify acronyms in text&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Stuart &lt;span

class=SpellE&gt;Yeates&lt;/span&gt;, David Bainbridge, Ian H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Text mining is

about looking for patterns in natural language text, and may be defined as the

process of &lt;span class=SpellE&gt;analyzing&lt;/span&gt; text to extract information from

it for particular purposes.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In previous

work, we claimed that compression is a key technology for text mining, and

backed this up with a study that showed how particular kinds of lexical

tokens—names, dates, locations, &lt;i style='mso-bidi-font-style:normal'&gt;etc.&lt;/i&gt;—can

be identified and located in running text, using compression models to provide

the leverage necessary to distinguish different token types (Witten &lt;i

style='mso-bidi-font-style:normal'&gt;et al.&lt;/i&gt;, 1999)&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/2&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Text &lt;span

class=SpellE&gt;categorization&lt;/span&gt; using compression models&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;Eibe&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt; Frank, Chang &lt;span class=SpellE&gt;Chui&lt;/span&gt;,

Ian H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Text &lt;span

class=SpellE&gt;categorization&lt;/span&gt;, or the assignment of natural language texts

to predefined categories based on their content, is of growing importance as

the volume of information available on the internet continues to overwhelm

us.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The use of predefined categories implies

a “supervised learning” approach to &lt;span class=SpellE&gt;categorization&lt;/span&gt;,

where already-classified articles – which effectively define the categories –

are used as “training data” to build a model that can be used for classifying

new articles that comprise the “test data”.&lt;span style='mso-spacerun:yes'&gt; 

&lt;/span&gt;This contrasts with “unsupervised” learning, where there is no training

data and clusters of like documents are sought amongst the test articles.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;With supervised learning, meaningful labels

(such as &lt;span class=SpellE&gt;keyphrases&lt;/span&gt;) are attached to the training

documents, and appropriate labels can be assigned automatically to test

documents depending on which category they fall into.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/3&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Reserved for

Sally Jo&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/4&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Interactive

machine learning—letting users build classifiers&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Malcolm Ware, &lt;span

class=SpellE&gt;Eibe&lt;/span&gt; Frank, Geoffrey Holmes, Mark Hall, Ian H. &lt;span

class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;According to

standard procedure, building a classifier is a fully automated process that

follows data preparation by a domain expert.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;In contrast, &amp;lt;I&amp;gt;interactive&amp;lt;/I&amp;gt;machine learning engages

users in actually generating the classifier themselves.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This offers a natural way of integrating

background knowledge into the &lt;span class=SpellE&gt;modeling&lt;/span&gt; stage—so long

as interactive tools can be designed that support efficient and effective

communication.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This paper shows that

appropriate techniques can empower users to create models that compete with

classifiers built by state-of-the-art learning algorithms.&lt;span

style='mso-spacerun:yes'&gt;  &lt;/span&gt;It demonstrates that users—even users who are

not domain experts—can often construct good classifiers, without any help from

a learning algorithm, using a simple two-dimensional visual interface.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Experiments demonstrate that, not

surprisingly, success hinges on the domain: if a few attributes can support

good predictions, users generate accurate classifiers, whereas domains with

many high-order attribute interactions &lt;span class=SpellE&gt;favor&lt;/span&gt; standard

machine learning techniques.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The future

challenge is to achieve a symbiosis between human user and machine learning

algorithm.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/5&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;KEA: Practical

automatic &lt;span class=SpellE&gt;keyphrase&lt;/span&gt; extraction&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Ian H. &lt;span

class=SpellE&gt;Witten&lt;/span&gt;, Gordon W. &lt;span class=SpellE&gt;Paynter&lt;/span&gt;, &lt;span

class=SpellE&gt;Eibe&lt;/span&gt; Frank, Carl &lt;span class=SpellE&gt;Gutwin&lt;/span&gt;, Craig G.

&lt;span class=SpellE&gt;Nevill&lt;/span&gt;-Manning&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;Keyphrases&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt; provide semantic metadata

that &lt;span class=SpellE&gt;summarize&lt;/span&gt; and &lt;span class=SpellE&gt;characterize&lt;/span&gt;

documents.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This paper describes &lt;span

class=SpellE&gt;Kea&lt;/span&gt;, an algorithm for automatically extracting &lt;span

class=SpellE&gt;keyphrases&lt;/span&gt; from text.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;&lt;span class=SpellE&gt;Kea&lt;/span&gt; identifies candidate &lt;span class=SpellE&gt;keyphrases&lt;/span&gt;

using lexical methods, calculates feature values for each candidate, and uses a

machine learning algorithm to predict which candidates are good &lt;span

class=SpellE&gt;keyphrases&lt;/span&gt;.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The

machine learning scheme first builds a prediction model using training

documents with known &lt;span class=SpellE&gt;keyphrases&lt;/span&gt;, and then uses the

model to find &lt;span class=SpellE&gt;keyphrases&lt;/span&gt; in new documents.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We use a large test corpus to evaluate &lt;span

class=SpellE&gt;Kea's&lt;/span&gt; effectiveness in terms of how many author-assigned &lt;span

class=SpellE&gt;keyphrases&lt;/span&gt; are correctly identified.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The system is simple, robust, and publicly

available.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/6&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;i style='mso-bidi-font-style:

normal'&gt;&lt;span lang=EN-GB style='font-family:Symbol;mso-ascii-font-family:&quot;Times New Roman&quot;;

mso-hansi-font-family:&quot;Times New Roman&quot;;mso-char-type:symbol;mso-symbol-font-family:

Symbol'&gt;&lt;span style='mso-char-type:symbol;mso-symbol-font-family:Symbol'&gt;m&lt;/span&gt;&lt;/span&gt;&lt;/i&gt;&lt;span

lang=EN-GB&gt;-Charts and Z:&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;&lt;span

class=SpellE&gt;hows&lt;/span&gt;, &lt;span class=SpellE&gt;whys&lt;/span&gt; and &lt;span

class=SpellE&gt;wherefores&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Greg Reeve,

Steve Reeves&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;In this paper we

show, by a series of examples, how the &lt;/span&gt;&lt;i style='mso-bidi-font-style:

normal'&gt;&lt;span lang=EN-GB style='font-family:Symbol;mso-ascii-font-family:&quot;Times New Roman&quot;;

mso-hansi-font-family:&quot;Times New Roman&quot;;mso-char-type:symbol;mso-symbol-font-family:

Symbol'&gt;&lt;span style='mso-char-type:symbol;mso-symbol-font-family:Symbol'&gt;m&lt;/span&gt;&lt;/span&gt;&lt;/i&gt;&lt;span

lang=EN-GB&gt;-chart formalism can be translated into Z.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We give reasons for why this is an

interesting and sensible thing to do and what it might be used for.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/7&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;One dimensional

non-uniform rational B-splines for animation control&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;Abdelaziz&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt; &lt;span class=SpellE&gt;Mahoui&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Most 3D

animation packages use graphical representations called motion graphs to

represent the variation in time of the motion parameters.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Many use two-dimensional B-splines as

animation curves because of their power to represent free-form curves.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In this project, we investigate the

possibility of using One-dimensional Non-Uniform Rational B-&lt;span class=SpellE&gt;Spline&lt;/span&gt;

(NURBS) curves for the interactive construction of animation control

curves.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;One-dimensional NURBS curves

present the potential of solving some problems encountered in motion graphs

when two-dimensional B-splines are used.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;The study focuses on the properties of One-dimensional NURBS

mathematical model.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It also investigates

the algorithms and shape modification tools devised for two-dimensional curves

and their port to the One-dimensional NURBS model.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It also looks at the issues related to the

user interface used to interactively modify the shape of the curves.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/8&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Correlation-based

feature selection of discrete and numeric class machine learning&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Mark A. Hall&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Algorithms for

feature selection fall into two broad categories:

&amp;lt;I&amp;gt;wrappers&amp;lt;/I&amp;gt;that use the learning algorithm itself to evaluate

the usefulness of features and &amp;lt;I&amp;gt;filters&amp;lt;/I&amp;gt;that evaluate features

according to heuristics based on general characteristics of the data.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;For application to large databases, filters

have proven to be more practical than wrappers because they are much

faster.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;However, most existing filter

algorithms only work with discrete classification problems.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This paper describes a fast,

correlation-based filter algorithm that can be applied to continuous and

discrete problems.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The algorithm often

out-performs the well-known &lt;span class=SpellE&gt;ReliefF&lt;/span&gt; attribute

estimator when used as a &lt;span class=SpellE&gt;preprocessing&lt;/span&gt; step for naïve

&lt;span class=SpellE&gt;Bayes&lt;/span&gt;, instance-based learning, decision trees,

locally weighted regression, and model trees.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;It performs more feature selection than &lt;span class=SpellE&gt;ReliefF&lt;/span&gt;

does-reducing the data dimensionality by fifty percent in most cases.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Also, decision and model trees built from the

&lt;span class=SpellE&gt;prepocessed&lt;/span&gt; data are often significantly smaller.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/9&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;A development

environment for predictive modelling in foods&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;G. Holmes, &lt;span

class=SpellE&gt;M.A.&lt;/span&gt; Hall&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;WEKA (Waikato

Environment for Knowledge Analysis) is a comprehensive suite of Java class

libraries that implement many state-of-the-art machine learning/data mining

algorithms.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Non-programmers interact

with the software via a user interface component called the Knowledge Explorer.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Applications

constructed from the WEKA class libraries can be run on any computer with a web

browsing capability, allowing users to apply machine learning techniques to

their own data regardless of computer platform.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;This paper describes the user interface component of the WEKA system in

reference to previous applications in the predictive &lt;span class=SpellE&gt;modeling&lt;/span&gt;

of foods.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/10&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Benchmarking

attribute selection techniques for data mining&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Mark A. Hall,

Geoffrey Holmes&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Data engineering

is generally considered to be a central issue in the development of data mining

applications.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The success of many

learning schemes, in their attempts to construct models of data, hinges on the

reliable identification of a small set of highly predictive attributes.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The inclusion of irrelevant, redundant and

noisy attributes in the model building process phase can result in poor

predictive performance and increased computation.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Attribute

selection generally involves a combination of search and attribute utility

estimation plus evaluation with respect to specific learning schemes.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This leads to a large number of possible

permutations and has led to a situation where very few benchmark studies have

been conducted.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

presents a benchmark comparison of several attribute selection methods.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;All the methods produce an attribute ranking,

a useful devise of isolating the individual merit of an attribute.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Attribute selection is achieved by

cross-validating the rankings with respect to a learning scheme to find the

best attributes.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Results are reported

for a selection of standard data sets and two learning schemes C4.5 and naïve &lt;span

class=SpellE&gt;Bayes&lt;/span&gt;.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/11&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Steve Reeves,

Greg Reeve&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;2000/12&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;Malika&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt; &lt;span class=SpellE&gt;Mahoui&lt;/span&gt;,

Sally Jo Cunningham&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Transaction logs

are invaluable sources of fine-grained information about users' search &lt;span

class=SpellE&gt;behavior&lt;/span&gt;.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This paper

compares the searching &lt;span class=SpellE&gt;behavior&lt;/span&gt; of users across two

WWW-accessible digital libraries: the New Zealand Digital Library's Computer

Science Technical Reports collection (CSTR), and the &lt;span class=SpellE&gt;Karlsruhe&lt;/span&gt;

Computer Science Bibliographies (CSBIB) collection.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Since the two collections are designed to

support the same type of users-researchers/students in computer science a

comparative log analysis is likely to uncover common searching preferences for

that user group.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The two collections

differ in their content, however; the CSTR indexes a full text collection,

while the CSBIB is primarily a bibliographic database.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Differences in searching &lt;span class=SpellE&gt;behavior&lt;/span&gt;

between the two systems may indicate the effect of differing search facilities

and content type.&lt;/span&gt;&lt;/p&gt;



























&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/1&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Lexical

attraction for text compression&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;Joscha&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt; Bach, Ian H. &lt;span

class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;New methods of

acquiring structural information in text documents may support better

compression by identifying an appropriate prediction context for each

symbol.&lt;span style='mso-spacerun:yes'&gt;  &lt;/span&gt;The method of “lexical

attraction” infers syntactic dependency structures from statistical analysis of

large corpora.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We describe the

generation of a lexical attraction model, discuss its application to text

compression, and explore its potential to outperform fixed-context models such

as word-level PPM.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Perhaps the most

exciting aspect of this work is the prospect of using compression as a metric

for structure discovery in text.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/2&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Generating rule

sets from model trees&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Geoffrey Holmes,

Mark Hall, &lt;span class=SpellE&gt;Eibe&lt;/span&gt; Frank&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Knowledge discovered

in a database must be represented in a form that is easy to understand.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Small, easy to interpret nuggets of knowledge

from data are one requirement and the ability to induce them from a variety of

data sources is a second.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The literature

is abound with classification algorithms, and in recent years with algorithms

for time sequence analysis, but relatively little has been published on

extracting meaningful information from problems involving continuous classes

(regression).&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Model

trees-decision trees with linear models at the leaf nodes-have recently emerged

as an accurate method for numeric prediction that produces understandable

models.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;However, it is well known that

decision lists-ordered sets of If-Then rules-have the potential to be more compact

and therefore more understandable than their tree counterparts.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;In this paper we

present an algorithm for inducing simple, yet accurate rule sets from model

trees.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The algorithm works by repeatedly

building model trees and selecting the best rule at each iteration.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It produces rule sets that are, on the whole,

as accurate but smaller than the model tree constructed from the entire &lt;span

class=SpellE&gt;dataset&lt;/span&gt;.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Experimental results for various heuristics which attempt to find a

compromise between rule accuracy and rule coverage are reported.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We also show empirically that our method

produces more accurate and smaller rule sets than the commercial

state-of-the-art rule learning system Cubist.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/3&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;A diagnostic

tool for tree based supervised classification learning algorithms&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Leonard &lt;span

class=SpellE&gt;Trigg&lt;/span&gt;, Geoffrey Holmes&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The process of

developing applications of machine learning and data mining that employ

supervised classification algorithms includes the important step of knowledge

verification.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Interpretable output is

presented to a user so that they can verify that the knowledge contained in the

output makes sense for the given application.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;As the development of an application is an iterative process it is quite

likely that a user would wish to compare models constructed at various times or

stages.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;One crucial

stage where comparison of models is important is when the accuracy of a model

is being estimated, typically using some form of cross-validation.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This stage is used to establish an estimate

of how well a model will perform on unseen data.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This is vital information to present to a

user, but it is also important to show the degree of variation between models

obtained from the entire &lt;span class=SpellE&gt;dataset&lt;/span&gt; and models obtained

during cross-validation.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In this way it

can be verified that the cross-validation models are at least structurally

aligned with the model garnered from the entire &lt;span class=SpellE&gt;dataset&lt;/span&gt;.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

presents a diagnostic tool for the comparison of tree-based supervised

classification models.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The method is

adapted from work on approximate tree matching and applied to decision

trees.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The tool is described together

with experimental results on standard &lt;span class=SpellE&gt;datasets&lt;/span&gt;.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/4&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Feature

selection for discrete and numeric class machine learning&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Mark A. Hall&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Algorithms for

feature selection fall into two broad categories:

&amp;lt;I&amp;gt;wrappers&amp;lt;/I&amp;gt;use the learning algorithm itself to evaluate the

usefulness of features, while &amp;lt;I&amp;gt;filters&amp;lt;/I&amp;gt;evaluate features

according to heuristics based on general characteristics of the data.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;For application to large databases, filters

have proven to be more practical than wrappers because they are much

faster.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;However, most existing filter

algorithms only work with discrete classification problems.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

describes a fast, correlation-based filter algorithm that can be applied to

continuous and discrete problems.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Experiments using the new method as a &lt;span class=SpellE&gt;preprocessing&lt;/span&gt;

step for naïve &lt;span class=SpellE&gt;Bayes&lt;/span&gt;, instance-based learning,

decision trees, locally weighted regression, and model trees show it to be an

effective feature selector- it reduces the data in dimensionality by more than

sixty percent in most cases without negatively affecting accuracy.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Also, decision and model trees built from the

pre-processed data are often significantly smaller.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/5&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Browsing tree

structures&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Mark &lt;span

class=SpellE&gt;Apperley&lt;/span&gt;, Robert &lt;span class=SpellE&gt;Spence&lt;/span&gt;, Stephen &lt;span

class=SpellE&gt;Hodge&lt;/span&gt;, Michael Chester&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Graphic

representations of tree structures are notoriously difficult to create,

display, and interpret, particularly when the volume of information they

contain, and hence the number of nodes, is large.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The problem of interactively browsing

information held in tree structures is examined, and the implementation of an

innovative tree browser described.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This

browser is based on distortion-oriented display techniques and intuitive direct

manipulation interaction.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The tree

layout is automatically generated, but the location and extent of detail shown

is controlled by the user.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It is

suggested that these techniques could be extended to the browsing of more

general networks.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/6&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Facilitating

multiple copy/past operations&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Mark &lt;span

class=SpellE&gt;Apperley&lt;/span&gt;, Jay Baker, Dale Fletcher, Bill Rogers&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Copy and paste,

or cut and paste, using a clipboard or paste buffer has long been the principle

facility provided to users for transferring data between and within GUI

applications.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We argue that this

mechanism can be clumsy in circumstances where several pieces of information

must be moved systematically.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In two

situations - extraction of data fields from unstructured data found in a

directed search process, and reorganisation of computer program source text -

we present alternative, more natural, user interface facilities to make the

task less onerous, and to provide improved visual feedback during the

operation.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;For the data

extraction task we introduce the Stretchable Selection Tool, a &lt;span

class=SpellE&gt;semi&lt;/span&gt;-transparent overlay augmenting the mouse pointer to

automate paste operations and provide information to prompt the user.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We describe a prototype implementation that

functions in a collaborative software environment, allowing users to &lt;span

class=SpellE&gt;cooperate&lt;/span&gt; on a multiple copy/paste operation.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;For text reorganisation, we present an

extension to &lt;span class=SpellE&gt;Emacs&lt;/span&gt;, providing similar functionality,

but without the collaborative features.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/7&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Automating

iterative tasks with programming by demonstration: a user evaluation&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Gordon W. &lt;span

class=SpellE&gt;Paynter&lt;/span&gt;, Ian H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Computer users

often face iterative tasks that cannot be automated using the tools and

aggregation techniques provided by their application program: they end up

performing the iteration by hand, repeating user interface actions over and

over again.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We have implemented an

agent, called Familiar, that can be taught to perform iterative tasks using

programming by demonstration (PBD).&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Unlike other PBD systems, it is domain independent and works with

unmodified, widely-used, applications in a popular operating system.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In a formal evaluation, we found that users

quickly learned to use the agent to automate iterative tasks.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Generally, the participants preferred to use

multiple selection where possible, but could and did use PBD in situations

involving iteration over many commands, or when other techniques were

unavailable.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/8&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;A survey of

software requirements specification practices in the New Zealand software

industry&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Lindsay Groves,

Ray &lt;span class=SpellE&gt;Nickson&lt;/span&gt;, Greg Reeve, Steve Reeves, Mark Utting&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We report on the

software development techniques used in the New Zealand software industry,

paying particular attention to requirements gathering.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We surveyed a selection of software companies

with a general questionnaire and then conducted in-depth interviews with four

companies.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Our results show a wide

variety in the kinds of companies undertaking software development, employing a

wide range of software development techniques.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Although our data are not sufficiently detailed to draw statistically

significant conclusions, it appears that larger software development groups

typically have more well-defined software development processes, spend

proportionally more time on requirements gathering, and follow more rigorous

testing regimes.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/9&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The LRU*WWW proxy

cache document replacement algorithm&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Chung-&lt;span

class=SpellE&gt;yi&lt;/span&gt; Chang, Tony &lt;span class=SpellE&gt;McGregor&lt;/span&gt;, Geoffrey

Holmes&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Obtaining good

performance from WWW proxy caches is critically dependent on the document

replacement policy used by the proxy.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;This paper validates the work of other authors by reproducing their

studies of proxy cache document replacement algorithms.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;From this basis a cross-trace study is

mounted.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This demonstrates that the

performance of most document replacement algorithms is dependent on the type of

workload that they are presented with.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Finally we propose a new algorithm, LRU*, that consistently performs

well across all our traces.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/10&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Reduced-error

pruning with significance tests&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;Eibe&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt; Frank, Ian H. &lt;span

class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;When building

classification models, it is common practice to prune them to counter spurious

effects of the training data: this often improves performance and reduces model

size.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;&amp;quot;Reduced-error pruning&amp;quot;

is a fast pruning procedure for decision trees that is known to produce small

and accurate trees.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Apart from the data

from which the tree is grown, it uses an independent &amp;quot;pruning&amp;quot; set,

and pruning decisions are based on the model's error rate on this fresh

data.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Recently it has been observed that

reduced-error pruning &lt;span class=SpellE&gt;overfits&lt;/span&gt; the pruning data,

producing unnecessarily large decision trees.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;This paper investigates whether standard statistical significance tests

can be used to counter this phenomenon.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The problem of &lt;span

class=SpellE&gt;overfitting&lt;/span&gt; to the pruning set highlights the need for

significance testing.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We investigate two

classes of test, &amp;quot;parametric&amp;quot; and &amp;quot;non-parametric.&amp;quot;&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The standard chi-squared statistic can be

used both in a parametric test and as the basis for a non-parametric

permutation test.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In both cases it is

necessary to select the significance level at which pruning is applied.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We show empirically that both versions of the

chi-squared test perform equally well if their significance levels are adjusted

appropriately.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Using a collection of

standard &lt;span class=SpellE&gt;datasets&lt;/span&gt;, we show that significance testing

improves on standard reduced error pruning if the significance level is

tailored to the particular &lt;span class=SpellE&gt;dataset&lt;/span&gt; at hand using

cross-validation, yielding consistently smaller trees that perform at least as

well and sometimes better.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/11&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;Weka&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt;: Practical machine learning

tools and techniques with Java implementations&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Ian H. &lt;span

class=SpellE&gt;Witten&lt;/span&gt;, &lt;span class=SpellE&gt;Eibe&lt;/span&gt; Frank, Len &lt;span

class=SpellE&gt;Trigg&lt;/span&gt;, Mark Hall, Geoffrey Holmes, Sally Jo Cunningham&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The Waikato

Environment for Knowledge Analysis (Weka) is a comprehensive suite of Java

class libraries that implement many state-of-the-art machine learning and data

mining algorithms.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;&lt;span class=SpellE&gt;Weka&lt;/span&gt;

is freely available on the &lt;span class=SpellE&gt;World-Wide&lt;/span&gt; Web and

accompanies a new text on data mining [1] which documents and fully explains

all the algorithms it contains.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Applications written using the &lt;span class=SpellE&gt;Weka&lt;/span&gt; class

libraries can be run on any computer with a Web browsing capability; this

allows users to apply machine learning techniques to their own data regardless

of computer platform.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/12&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Pace Regression&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Yong Wang, Ian

H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

articulates a new method of linear regression, “pace regression”, that

addresses many drawbacks of standard regression reported in the

literature—particularly the subset selection problem.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Pace regression improves on classical ordinary

least squares (OLS) regression by evaluating the effect of each variable and

using a clustering analysis to improve the statistical basis for estimating

their contribution to the overall regression.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;As well as outperforming OLS, it also outperforms—in a remarkably

general sense—other linear &lt;span class=SpellE&gt;modeling&lt;/span&gt; techniques in the

literature, including subset selection procedures, which seek a reduction in

dimensionality that falls out as a natural &lt;span class=SpellE&gt;byproduct&lt;/span&gt;

of pace regression.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The paper defines

six procedures that share the fundamental idea of pace regression, all of which

are theoretically justified in terms of asymptotic performance.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Experiments confirm the performance

improvement over other techniques.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/13&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;A

compression-based algorithm for Chinese word segmentation&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;W.J. &lt;span

class=SpellE&gt;Teahan&lt;/span&gt;, &lt;span class=SpellE&gt;Yingying&lt;/span&gt; Wen, &lt;span

class=SpellE&gt;Rodger&lt;/span&gt; &lt;span class=SpellE&gt;McNab&lt;/span&gt;, Ian H. &lt;span

class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The Chinese

language is written without using spaces or other word delimiters.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Although a text may be thought of as a

corresponding sequence of words, there is considerable ambiguity in the

placement of boundaries.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Interpreting a

text as a sequence of words is beneficial for some information retrieval and

storage tasks: for example, full-text search, word-based compression, and &lt;span

class=SpellE&gt;keyphrase&lt;/span&gt; extraction.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We describe a

scheme that infers appropriate positions for word boundaries using an adaptive

language model that is standard in text compression.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It is trained on a corpus of pre-segmented

text, and when applied to new text, interpolates word boundaries so as to &lt;span

class=SpellE&gt;maximize&lt;/span&gt; the compression obtained.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This simple and general method performs well

with respect to &lt;span class=SpellE&gt;specialized&lt;/span&gt; schemes for Chinese

language segmentation.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/14&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Clustering with

finite data from &lt;span class=SpellE&gt;semi&lt;/span&gt;-parametric mixture

distributions&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Yong Wang, Ian

H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Existing

clustering methods for the &lt;span class=SpellE&gt;semi&lt;/span&gt;-parametric mixture

distribution perform well as the volume of data increases.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;However, they all suffer from a serious

drawback in finite-data situations: small outlying groups of data points can be

completely ignored in the clusters that are produced, no matter how far away

they lie from the major clusters.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This

can result in unbounded loss if the loss function is sensitive to the distance

between clusters.&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

proposes a new distance-based clustering method that overcomes the problem by

avoiding global constraints.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Experimental results illustrate its superiority to existing methods when

small clusters are present in finite data sets; they also suggest that it is

more accurate and stable than other methods even when there are no small

clusters.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/15&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;99/16&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The &lt;span

class=SpellE&gt;Niupepa&lt;/span&gt; Collection:&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Opening

the blinds on a window to the past&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;Te&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt; &lt;span class=SpellE&gt;Taka&lt;/span&gt; &lt;span

class=SpellE&gt;Keegan&lt;/span&gt;, Sally Jo Cunningham, Mark &lt;span class=SpellE&gt;Apperley&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

describes the building of a digital library collection of historic

newspapers.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The newspapers (&lt;span

class=SpellE&gt;&lt;i style='mso-bidi-font-style:normal'&gt;Niupepa&lt;/i&gt;&lt;/span&gt; in &lt;span

class=SpellE&gt;Maori&lt;/span&gt;), which were published in New Zealand during the

period 1842 to 1933, form a unique historical record of the &lt;span class=SpellE&gt;Maori&lt;/span&gt;

language, and of events from an historical perspective.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Images of these newspapers have been

converted to digital form, electronic text extracted from these, and the

collection is now being made available over the Internet as a part of the New

Zealand Digital Library (NZDL) project at the University of Waikato.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/1&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Boosting trees

for cost-sensitive classifications&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Kai &lt;span

class=SpellE&gt;Ming&lt;/span&gt; Ting, &lt;span class=SpellE&gt;Zijian&lt;/span&gt; &lt;span

class=SpellE&gt;Zheng&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

explores two boosting techniques for cost-sensitive tree classification in the

situation where misclassification costs change very often.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Ideally, one would like to have only one

induction, and use the induced model for different misclassification

costs.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Thus, it demands robustness of

the induced model against cost changes.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Combining multiple trees gives robust predictions against this

change.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We demonstrate that ordinary

boosting combined with the minimum expected cost criterion to select the

prediction class is a good solution under this situation.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We also introduce a variant of the ordinary

boosting procedure which &lt;span class=SpellE&gt;utilizes&lt;/span&gt; the cost

information during training.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We show

that the proposed technique performs better than the ordinary boosting in terms

of misclassification cost.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;However, this

technique requires to induce a set of new trees every time the cost

changes.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Our empirical investigation

also reveals some interesting &lt;span class=SpellE&gt;behavior&lt;/span&gt; of boosting

decision trees for cost-sensitive classification.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/2&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Generating

accurate rule sets without global &lt;span class=SpellE&gt;optimization&lt;/span&gt; &lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;Eibe&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt; Frank, Ian H. &lt;span

class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The two dominant

schemes for rule-learning, C4.5 and RIPPER, both operate in two stages.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;First they induce an initial rule set and

then they refine it using a rather complex &lt;span class=SpellE&gt;optimization&lt;/span&gt;

stage that discards (C4.5) or adjusts (RIPPER) individual rules to make them

work better together.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In contrast, this

paper shows how good rule sets can be learned one rule at a time, without any

need for global &lt;span class=SpellE&gt;optimization&lt;/span&gt;.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We present an algorithm for inferring rules

by repeatedly generating partial decision trees, thus combining the two major

paradigms for rule generation-creating rules from decision trees and the

separate-and-conquer rule-learning technique.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;The algorithm is straightforward and elegant: despite this, experiments

on standard &lt;span class=SpellE&gt;datasets&lt;/span&gt; show that it produces rule sets

that are as accurate as and of similar size to those generated by C4.5, and

more accurate than &lt;span class=SpellE&gt;RIPPER's&lt;/span&gt;.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Moreover, it operates efficiently, and

because it avoids &lt;span class=SpellE&gt;postprocessing&lt;/span&gt;, does not suffer the

extremely slow performance on pathological example sets for which the C4.5

method has been &lt;span class=SpellE&gt;criticized&lt;/span&gt;.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/3&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;VQuery&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt;: a graphical user interface

for Boolean query Specification and dynamic result preview&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Steve Jones&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Textual query

languages based on Boolean logic are common amongst the search facilities of

on-line information repositories.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;However, there is evidence to suggest that the syntactic and semantic

demands of such languages lead to user errors and adversely affect the time

that it takes users to form queries.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Additionally, users are faced with user interfaces to these repositories

which are unresponsive and uninformative, and consequently fail to support

effective query refinement.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We suggest

that graphical query languages, particularly Venn-like diagrams, provide a

natural medium for Boolean query specification which overcomes the problems of

textual query languages.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Also, dynamic

result previews can be seamlessly integrated with graphical query specification

to increase the effectiveness of query refinements.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We describe &lt;span class=SpellE&gt;VQuery&lt;/span&gt;,

a query interface to the New Zealand Digital Library which exploits querying by

Venn diagrams and integrated query result previews.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/4&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Revising

&amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt;: semantics and logic&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Martin C. &lt;span

class=SpellE&gt;Henson&lt;/span&gt;, Steve Reeves&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We introduce a

simple specification logic &amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt;c comprising a logic and

semantics (in &amp;lt;I&amp;gt;ZF&amp;lt;/I&amp;gt; set theory).&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We then provide an interpretation for (a

rational reconstruction of) the specification language &amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt;

within &amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt;c.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;As a

result we obtain a sound logic for &amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt;, including the schema

calculus.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;A consequence of our

formalisation is a critique of a number of concepts used in

&amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt;.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We demonstrate

that the complications and confusions which these concepts introduce can be avoided

without compromising &lt;span class=SpellE&gt;expressibility&lt;/span&gt;.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/5&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;A logic for the

schema calculus&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Martin C. &lt;span

class=SpellE&gt;Henson&lt;/span&gt;, Steve Reeves&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;In this paper we

introduce and investigate a logic for the schema calculus of

&amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt;.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The schema

calculus is arguably the reason for &amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt;’s popularity but so

far no true calculus (a sound system of rules for reasoning about schema

expressions) has been given.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Presentations thus far have either failed to provide a calculus (e.g.

the draft standard [3]) or have fallen back on informal descriptions at a

syntactic level (most text books e.g. [7[).&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Once the calculus is established we introduce a derived &lt;span

class=SpellE&gt;equational&lt;/span&gt; logic which enables us to formalise properly the

informal notations of schema expression equality to be found in the literature.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/6&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;New foundations

for &amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Martin C. &lt;span

class=SpellE&gt;Henson&lt;/span&gt;, Steve Reeves&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We provide a

constructive and &lt;span class=SpellE&gt;intensional&lt;/span&gt; interpretation for the

specification language &amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt; in a theory of operations and kinds

&amp;lt;I&amp;gt;T&amp;lt;/I&amp;gt;.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The motivation is

to facilitate the development of an integrated approach to program

construction.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We illustrate the new

foundations for &amp;lt;I&amp;gt;Z&amp;lt;/I&amp;gt; with examples.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/7&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Predicting apple

bruising relationships using machine learning&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;G. Holmes, S.J.

Cunningham, B.T. &lt;span class=SpellE&gt;Dela&lt;/span&gt; Rue, &lt;span class=SpellE&gt;A.F.&lt;/span&gt;

&lt;span class=SpellE&gt;Bollen&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoBodyText&gt;&lt;span lang=EN-US&gt;Many models have been used to describe

the influence of internal or external factors on apple bruising.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Few of these have addressed the application

of derived relationships to the evaluation of commercial operations.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;From an industry perspective, a model must

enable fruit to be rejected on the basis of a commercially significant bruise

and must also accurately quantify the effects of various combinations of input

features (such as &lt;span class=SpellE&gt;cultivar&lt;/span&gt;, maturity, size, and so

on) on bruise prediction.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Input features

must in turn have characteristics which are measurable commercially; for

example, the measure of force should be impact energy rather than energy

absorbed.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Further, as the commercial

criteria for acceptable damage levels change, the model should be versatile

enough to regenerate new bruise thresholds from existing data.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Machine learning

is a burgeoning technology with a vast range of potential applications

particularly in agriculture where large amounts of data can be readily

collected [1].&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The main advantage of

using a machine learning method in an application is that the models built for

prediction can be viewed and understood by the owner of the data who is in a

position to determine the usefulness of the model, an essential component in a

commercial environment.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/8&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;An evaluation of

passage-level indexing strategies for a technical report archive&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Michael Williams&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Past research

has shown that using evidence from document passages rather than complete

documents is an effective way of improving the precision of full-text database

searches.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;However, passage-level

indexing has yet to be widely adopted for commercial or online databases.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

reports on experiments designed to test the efficacy of passage-level indexing

with a particular collection of a full-text online database, the New Zealand

Digital Library.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Discourse passages and

word-window passages are used for the indexing process.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Both ranked and Boolean searching are used to

test the resulting indexes.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Overlapping

window passages are shown to offer the best retrieval performance with both

ranked and Boolean queries.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Modifications may be necessary to the term weighting methodology in

order to ensure optimal ranked query performance.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/9&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Managing

multiple collections, multiple languages, and multiple media in a distributed

digital library&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Ian H. &lt;span

class=SpellE&gt;Witten&lt;/span&gt;, &lt;span class=SpellE&gt;Rodger&lt;/span&gt; &lt;span

class=SpellE&gt;McNab&lt;/span&gt;, Steve Jones, Sally Jo Cunningham, David Bainbridge,

Mark &lt;span class=SpellE&gt;Apperley&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Managing the &lt;span

class=SpellE&gt;organizational&lt;/span&gt; and software complexity of a comprehensive

digital library presents a significant challenge.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Different library collections each have their

own distinctive features.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Different

presentation languages have structural implications such as left-to-right

writing order and text-only interfaces for the visually impaired.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Different media involve different file

formats, and-more importantly-radically different search strategies are

required for non-textual media.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In a

distributed library, new collections can appear asynchronously on servers in

different parts of the world.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;And as

searching interfaces mature from the command-line era exemplified by current

Web search engines into the age of reactive visual interfaces, experimental new

interfaces must be developed, supported, and tested.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This paper describes our experience, gained

from operating a substantial digital library service over several years, in

solving these problems by designing an appropriate software architecture.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/10&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Experiences with

a weighted decision tree learner&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;John G. &lt;span

class=SpellE&gt;Cleary&lt;/span&gt;, Leonard E. &lt;span class=SpellE&gt;Trigg&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoBodyText&gt;&lt;span lang=EN-US&gt;Machine learning algorithms for inferring

decision trees typically choose a single “best” tree to describe the training

data.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Recent research has shown that

classification performance can be significantly improved by voting predictions

of multiple, independently produced decision trees.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This paper describes an algorithm, OB1, that

makes a weighted sum over many possible models.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;We describe one instance of OB1, that includes &amp;lt;I&amp;gt;all&amp;lt;/I&amp;gt;

possible decision trees as well as naïve &lt;span class=SpellE&gt;Bayesian&lt;/span&gt;

models.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;OB1 is compared with a number of

other decision tree and instance based learning &lt;span class=SpellE&gt;alogrithms&lt;/span&gt;

on some of the data sets from the UCI repository.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Both an information gain and an accuracy

measure are used for the comparison.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;On

the information gain measure OB1 performs significantly better than all the

other algorithms.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;On the accuracy

measure it is significantly better than all the algorithms except naïve &lt;span

class=SpellE&gt;Bayes&lt;/span&gt; which performs comparably to OB1.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/11&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;An entropy gain

measure of numeric prediction performance&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Leonard &lt;span

class=SpellE&gt;Trigg&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Categorical

classifier performance is typically evaluated with respect to error rate,

expressed as a percentage of test instances that were not correctly

classified.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;When a classifier produces

multiple classifications for a test instance, the prediction is counted as

incorrect (even if the correct class was one of the predictions).&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Although commonly used in the literature,

error rate is a coarse measure of classifier performance, as it is based only

on a single prediction offered for a test instance.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Since many classifiers can produce a class

distribution as a prediction, we should use this to provide a better measure of

how much information the classifier is extracting from the domain.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Numeric

classifiers are a relatively new development in machine learning, and as such

there is no single performance measure that has become standard.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Typically these machine learning schemes

predict a single real number for each test instance, and the error between the

predicted and actual value is used to calculate a myriad of performance

measures such as correlation coefficient, root mean squared error, mean

absolute error, relative absolute error, and root relative squared error.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;With so many performance measures it is

difficult to establish an overall performance evaluation.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The next section

describes a performance measure for machine learning schemes that attempts to

overcome the problems with current measures.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;In addition, the same evaluation measure is used for categorical and

numeric classifier.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/12&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Proceedings of

CBISE ’98 CaiSE*98 Workshop on Component Based Information Systems Engineering&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Edited by John &lt;span

class=SpellE&gt;Grundy&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoBodyText&gt;&lt;span lang=EN-US&gt;Component-based information systems

development is an area of research and practice of increasing importance.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Information Systems developers have &lt;span

class=SpellE&gt;realised&lt;/span&gt; that traditional approaches to IS engineering

produce monolithic, difficult to maintain, difficult to reuse systems.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In contrast, the use of software components,

which embody data, functionality and well-specified and understood interfaces,

makes interoperable, distributed and highly reusable IS components

feasible.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Component-based approaches to

IS engineering can be used at strategic and &lt;span class=SpellE&gt;organisational&lt;/span&gt;

levels, to model business processes and whole IS architectures, in development

methods which &lt;span class=SpellE&gt;utilise&lt;/span&gt; component-based models during

analysis and design, and in system implementation.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Reusable components can allow end users to

compose and configure their own Information Systems, possibly from a range of

suppliers, and to more tightly couple their &lt;span class=SpellE&gt;organisational&lt;/span&gt;

&lt;span class=SpellE&gt;workflows&lt;/span&gt; with their IS support.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This workshop

proceedings contains a range of papers addressing one or more of the above

issues relating to the use of component models for IS development.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;All of these papers were refereed by at least

two members of an international workshop committee comprising industry and

academic researchers and users of component technologies.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Strategic uses of components are addressed in

the first three papers, while the following three address uses of components for

systems design and workflow management.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Systems development using components, and the provision of environments

for component management are addressed in the following group of five

papers.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The last three papers in this

proceedings address component management and analysis techniques.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;All of these

papers provide new insights into the many&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;varied uses of component technology for IS engineering.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;I hope you find them as interesting and

useful as I have when collating this proceedings and organising the workshop.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/13&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;An analysis of

usage of a digital library&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Steve Jones,

Sally Jo Cunningham, &lt;span class=SpellE&gt;Rodger&lt;/span&gt; &lt;span class=SpellE&gt;McNab&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;As experimental

digital library &lt;span class=SpellE&gt;testbeds&lt;/span&gt; gain wider acceptance and

develop significant user bases, it becomes important to investigate the ways in

which users interact with the systems in practice.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Transaction logs are one source of usage

information, and the information on user behaviour can be culled from them both

automatically (through calculation of summary statistics) and manually (by

examining query strings for semantic clues on search motivations and searching

strategy).&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We conduct a transaction log

analysis on user activity in the Computer Science Technical Reports Collection

of the New Zealand Digital Library, and report insights gained and identify

resulting search interface design issues.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/14&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Measuring ATM

traffic: final report for New Zealand Telecom&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;John &lt;span

class=SpellE&gt;Cleary&lt;/span&gt;, Ian Graham, Murray Pearson, Tony &lt;span

class=SpellE&gt;McGregor&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The report

describes the development of a low-cost ATM monitoring system, hosted by a

standard PC.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The monitor can be used

remotely returning information on ATM traffic flows to a central site.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The monitor is interfaces to a GPS timing

receiver, which provides an absolute time accuracy of better than 1 &lt;span

class=SpellE&gt;usec&lt;/span&gt;.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;By monitoring

the same traffic flow at different points in a network it is possible to

measure cell delay and delay variation in real time, and with existing

traffic.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The monitoring system

characterises cells by a CRC calculated over the cell payload, thus special

measurement cells are not required.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Delays in both local area and wide-area networks have been measured

using this system.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It is possible to

measure delay in a network that is not end-to-end ATM, as long as some cells

remain identical at the entry and exit points.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Examples are given of traffic and delay measurements in both wide and

local area network systems, including delays measured over the Internet from

Canada to New Zealand.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/15&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Despite its

simplicity, the naïve &lt;span class=SpellE&gt;Bayes&lt;/span&gt; learning scheme performs

well on most classification tasks, and is often significantly more accurate

than more sophisticated methods.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Although the probability estimates that it produces can be inaccurate,

it often assigns maximum probability to the correct class.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This suggests that its good performance might

be restricted to situations where the output is categorical.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It is therefore interesting to see how it

performs in domains where the predicted value is numeric, because in this case,

predictions are more sensitive to inaccurate probability estimates.&amp;lt;P&amp;gt;&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper shows

how to apply the naïve &lt;span class=SpellE&gt;Bayes&lt;/span&gt; methodology to numeric

prediction (i.e. regression) tasks, and compares it to linear regression,

instance-based learning, and a method that produces “model trees”-decision

trees with linear regression functions at the leaves.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Although we exhibit an artificial &lt;span

class=SpellE&gt;dataset&lt;/span&gt; for which naïve &lt;span class=SpellE&gt;Bayes&lt;/span&gt; is

the method of choice, on real-world &lt;span class=SpellE&gt;datasets&lt;/span&gt; it is

almost uniformly worse than model trees.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;The comparison with linear regression depends on the error measure: for

one measure naïve &lt;span class=SpellE&gt;Bayes&lt;/span&gt; performs similarly, for

another it is worse.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Compared to

instance-based learning, it performs similarly with respect to both

measures.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;These results indicate that

the simplistic statistical assumption that naïve &lt;span class=SpellE&gt;Bayes&lt;/span&gt;

makes is indeed more restrictive for regression than for classification.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/16&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Link as you

type: using key phrases for automated dynamic link generation&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Steve Jones&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;When documents

are collected together from diverse sources they are unlikely to contain useful

hypertext links to support browsing amongst them.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;For large collections of thousands of

documents it is prohibitively resource intensive to manually insert links into

each document.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Users of such collections

may wish to relate documents within them to text that they are themselves

generating.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This process, often

involving keyword searching, distracts from the authoring process and results

in material related to query terms but not necessarily to the author’s

document.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Query terms that are effective

in one collection might not be so in another.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;We have developed &lt;span class=SpellE&gt;Phrasier&lt;/span&gt;, a system that

integrates authoring (of text and hyperlinks), browsing, querying and reading

in support of information retrieval activities.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;&lt;span class=SpellE&gt;Phrasier&lt;/span&gt; exploits key phrases which are

automatically extracted from documents in a collection, and uses them as link

anchors and to identify candidate destinations for hyperlinks.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This system suggests links into existing

collections for purposes of authoring and retrieval of related information,

creates links between documents in a collection and provides supportive

document and link overviews.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/17&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Melody based

tune retrieval over the World Wide Web&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;David

Bainbridge, &lt;span class=SpellE&gt;Rodger&lt;/span&gt; J. &lt;span class=SpellE&gt;McNab&lt;/span&gt;,

Lloyd A. Smith&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;In this paper we

describe the steps taken to develop a Web-based version of an existing

stand-alone, single-user digital library application for &lt;span class=SpellE&gt;melodical&lt;/span&gt;

searching of a collection of music.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;For

the three key components: input, searching, and output, we assess the

suitability of various Web-based strategies that deal with the now distributed

software architecture and explain the decisions we made.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The resulting melody indexing service, known

as MELDEX, has been in operation for one year, and the feed-back we have

received has been &lt;span class=SpellE&gt;favorable&lt;/span&gt;.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;98/18&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Making oral

history accessible over the World Wide Web&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;David

Bainbridge, Sally Jo Cunningham&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We describe a

multimedia, WWW-based oral history collection constructed from off-the-shelf or

publicly available software.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The source

materials for the collection include audio tapes of interviews and summary

transcripts of each interview, as well as photographs illustrating episodes

mentioned in the tapes.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Sections of the

transcripts are manually matched to associated segments of the tapes, and the

tapes are &lt;span class=SpellE&gt;digitized&lt;/span&gt;.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Users search a full-text retrieval system based on the text transcripts

to retrieve relevant transcript sections and their associated audio recordings

and photographs.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It is also possible to

search for photos by matching text queries against text descriptions of the

photos in the collection, where the located photos link back to their

respective interview transcript and audio recordings.&lt;/span&gt;&lt;/p&gt;



















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;b style='mso-bidi-font-weight:

normal'&gt;&lt;span lang=EN-GB&gt;1997&lt;o:p&gt;&lt;/o:p&gt;&lt;/span&gt;&lt;/b&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/1&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;A dynamic and

flexible representation of social relationships in CSCW&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Steve Jones,

Steve Marsh&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;CSCW system

designers lack effective support in addressing the social issues and

interpersonal relationships which are linked with the use of CSCW systems.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We present a formal description of trust to

support CSCW system designers in considering the social aspects of group work,

embedding those considerations in systems and analysing computer supported

group processes.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We argue that

trust is a critical aspect in group work, and describe what we consider to be

the building blocks of trust.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We then

present a formal notation for the building blocks, their use in reasoning about

social interactions and how they are amended over time.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We then consider

how the formalism may be used in practice, and present some insights from

initial analysis of the behaviour of the formalism.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This is followed by a description of possible

amendments and extensions to the formalism.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;We conclude that it is possible to formalise a notion of trust and to

model the formalisation by a computational mechanism.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/2&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Design issues

for World Wide Web navigation visualisation tools&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Andy &lt;span

class=SpellE&gt;Cockburn&lt;/span&gt;, Steve Jones&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The World Wide

Web (WWW) is a successful hypermedia information space used by millions of

people, yet it suffers from many deficiencies and problems in support for

navigation around its vast information space.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;In this paper we identify the origins of these navigation problems,

namely WWW browser design, WWW page design, and WWW page description

languages.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Regardless of their origins,

these problems are eventually represented to the user at the browser’s user

interface.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;To help overcome these

problems, many tools are being developed which allow users to visualise WWW

subspaces.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We identify five key issues

in the design and functionality of these visualisation systems: characteristics

of the visual representation, the scope of the subspace representation, the

mechanisms for generating the visualisation, the degree of browser

independence, and the navigation support facilities.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We provide a critical review of the diverse

range of WWW visualisation tools with respect to these issues.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/3&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Stacked &lt;span

class=SpellE&gt;generalization&lt;/span&gt;:&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;when

does it work?&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Kai &lt;span

class=SpellE&gt;Ming&lt;/span&gt; Ting, Ian H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Stacked &lt;span

class=SpellE&gt;generalization&lt;/span&gt; is a general method of using a high-level

model to combine lower-level models to achieve greater predictive

accuracy.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In this paper we address two

crucial issues which have been considered to be a 'black art' in classification

tasks ever since the introduction of stacked &lt;span class=SpellE&gt;generalization&lt;/span&gt;

in 1992 by &lt;span class=SpellE&gt;Wolpert&lt;/span&gt;: the type of &lt;span class=SpellE&gt;generalizer&lt;/span&gt;

that is suitable to derive the higher-level model, and the kind of attributes

that should be used as its input. &lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We demonstrate

the effectiveness of stacked &lt;span class=SpellE&gt;generalization&lt;/span&gt; for

combining three different types of learning algorithms, and also for combining

models of the same type derived from a single learning algorithm in a

multiple-data-batches scenario.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We also

compare the performance of stacked &lt;span class=SpellE&gt;generalization&lt;/span&gt;

with published results arcing and bagging.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/4&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Browsing in

digital libraries:&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;a phrase-based

approach&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Craig &lt;span

class=SpellE&gt;Nevill&lt;/span&gt;-Manning, Ian H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;,

Gordon W. &lt;span class=SpellE&gt;Paynter&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;A key question

for digital libraries is this: how should one go about becoming familiar with a

digital collection, as opposed to a physical one?&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Digital collections generally present an

appearance which is extremely opaque-a screen, typically a Web page, with no

indication of what, or how much, lies beyond: whether a carefully-selected

collection or a morass of worthless ephemera; whether half a dozen documents or

many millions.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;At least physical

collections occupy physical space, present a physical appearance, and exhibit

tangible physical &lt;span class=SpellE&gt;organization&lt;/span&gt;.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;When standing on the threshold of a large

library one gains a sense of presence and permanence that reflects the care

taken in building and maintaining the collection inside.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;No-one could confuse it with a

dung-heap!&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Yet in the digital world the

difference is not so palpable.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/5&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;A graphical

notation for the design of information visualisations&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Matthew C. &lt;span

class=SpellE&gt;Humphrey&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Visualisations

are coherent, graphical expressions of complex information that enhance people’s

ability to communicate and reason about that information.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Yet despite the importance of visualisations

in helping people to understand and solve a wide variety of problems, there is

a dearth of formal tools and methods for discussing, describing and designing

them.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Although simple visualisations,

such as bar charts and &lt;span class=SpellE&gt;scatterplots&lt;/span&gt;, are easily

produced by modern interactive software, novel visualisations of multivariate, &lt;span

class=SpellE&gt;multirelational&lt;/span&gt; data must be expressed in a programming

language.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The Relational Visualisation

Notation is a new, graphical language for designing such highly expressive

visualisations that does not use programming constructs.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Instead, the notation is based on relational

algebra, which is widely used in database query languages, and it is supported

by a suite of direct manipulation tools.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;This article presents the notation and examines the designs of some

interesting visualisations.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/6&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Applications of

machine learning in information retrieval&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Sally Jo

Cunningham, James &lt;span class=SpellE&gt;Littin&lt;/span&gt;, Ian H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Information

retrieval systems provide access to collections of thousands, or millions, of

documents, from which, by providing an appropriate description, users can

recover any one.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Typically, users &lt;span

class=SpellE&gt;iteratively&lt;/span&gt; refine the descriptions they provide to satisfy

their needs, and retrieval systems can &lt;span class=SpellE&gt;utilize&lt;/span&gt; user

feedback on selected documents to indicate the accuracy of the description at

any stage.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The style of description

required from the user, and the way it is employed to search the document

database, are consequences of the indexing method used for the collection.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The index may take different forms, from

storing keywords with links to individual documents, to clustering documents

under related topics.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/7&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Computer

concepts without computers:&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;a first

course in computer science&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Geoffrey Holmes,

Tony C. Smith, William J. Rogers&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;While some

institutions seek to make CS1 curricula more enjoyable by incorporating

specialised educational software [1] or by setting more enjoyable programming

assignments [2], we have joined the growing number of Computer Science

departments that seek to improve the quality of the CS1 experience by focusing

student attention away from the computer monitor [3,4].&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Sophisticated computing concepts usually

reserved for senior level courses are presented in a &amp;lt;I&amp;gt;popular

science&amp;lt;/I&amp;gt; manner, and given equal time alongside the essential

introductory programming material.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;By

exposing students to a broad range of specific computational problems we

endeavour to make the introductory course more interesting and enjoyable, and

instil in students a sense of vision for areas they might specialise in as

computing majors.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/8&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;A sight-singing

tutor&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Lloyd A. Smith, &lt;span

class=SpellE&gt;Rodger&lt;/span&gt; J. &lt;span class=SpellE&gt;McNab&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

describes a computer program designed to aid its users in learning to

sight-sing.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Sight-singing-the ability to

sing music from a score without prior study-is an important skill for musicians

and holds a central place in most university music curricula.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Its importance to vocalists is obvious; it is

also an important skill for instrumentalists and conductors because it develops

the aural imagination necessary to judge how the music should sound, when

played (&lt;span class=SpellE&gt;Benward&lt;/span&gt; and Carr 1991).&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Furthermore, it is an important skill for

amateur musicians, who can save a great deal of rehearsal time through an

ability to sing music at sight.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/9&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Stacking bagged

and &lt;span class=SpellE&gt;dagged&lt;/span&gt; models&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Kai &lt;span

class=SpellE&gt;Ming&lt;/span&gt; Ting, I.H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;In this paper,

we investigate the method of &lt;i style='mso-bidi-font-style:normal'&gt;stacked &lt;span

class=SpellE&gt;generalization&lt;/span&gt;&lt;/i&gt; in combining models derived from

different subsets of a training &lt;span class=SpellE&gt;dataset&lt;/span&gt; by a single

learning algorithm, as well as different algorithms.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The simplest way to combine predictions from

competing models is majority vote, and the effect of the sampling regime used

to generate training subsets has already been studied in this context-when

bootstrap samples are used the method is called &lt;i style='mso-bidi-font-style:

normal'&gt;bagging&lt;/i&gt;, and for disjoint samples we call it &lt;span class=SpellE&gt;&lt;i

style='mso-bidi-font-style:normal'&gt;dagging&lt;/i&gt;&lt;/span&gt;.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This paper extends these studies to stacked &lt;span

class=SpellE&gt;generalization&lt;/span&gt;, where a learning algorithm is employed to combine

the models.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This yields new methods

dubbed &lt;i style='mso-bidi-font-style:normal'&gt;bag-stacking&lt;/i&gt; and &lt;span

class=SpellE&gt;&lt;i style='mso-bidi-font-style:normal'&gt;dag&lt;/i&gt;&lt;/span&gt;&lt;i

style='mso-bidi-font-style:normal'&gt;-stacking&lt;/i&gt;.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We demonstrate

that bag-stacking and &lt;span class=SpellE&gt;dag&lt;/span&gt;-stacking can be effective

for classification tasks even when the training samples cover just a small

fraction of the full &lt;span class=SpellE&gt;dataset&lt;/span&gt;.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In contrast to earlier bagging results, we

show that bagging and bag-stacking work for stable as well as unstable learning

algorithms, as do &lt;span class=SpellE&gt;dagging&lt;/span&gt; and &lt;span class=SpellE&gt;dag&lt;/span&gt;-stacking.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We find that bag-stacking (&lt;span

class=SpellE&gt;dag&lt;/span&gt;-stacking) almost always has higher predictive accuracy

than bagging (&lt;span class=SpellE&gt;dagging&lt;/span&gt;), and we also show that

bag-stacking models derived using two different algorithms is more effective

than bagging.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/10&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Extracting text

from Postscript&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Craig &lt;span

class=SpellE&gt;Nevill&lt;/span&gt;-Manning, Todd Reed, Ian H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We show how to

extract plain text from PostScript files. A textual scan is inadequate because

PostScript interpreters can generate characters on the page that do not appear

in the source file. Furthermore, word and line breaks are implicit in the

graphical rendition, and must be inferred from the positioning of word

fragments. We present a robust technique for extracting text and &lt;span

class=SpellE&gt;recognizing&lt;/span&gt; words and paragraphs. The method uses a

standard PostScript interpreter but redefines several PostScript operators, and

simple heuristics are employed to locate word and line breaks. The scheme has

been used to create a full-text index, and plain-text versions, of 40,000

technical reports (34 &lt;span class=SpellE&gt;Gbyte&lt;/span&gt; of PostScript). Other

text-extraction systems are reviewed: none offer the same combination of

robustness and simplicity.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/11&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Gathering and

indexing rich fragments of the World Wide Web&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Geoffrey Holmes,

William J Rogers&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;While the World

Wide Web (WWW) is an attractive option as a resource for teaching and research

it does have some undesirable features. The cost of allowing students unlimited

access can be high-both in money and time; students may become addicted to

'surfing' the web-exploring purely for entertainment-and jeopardise their

studies. Students are likely to discover undesirable material because large

scale search engines index sites regardless of their merit. Finally, the

explosive growth of WWW usage means that servers and networks are often

overloaded, to the extent that a student may gain a very negative view of the

technology.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We have developed

a piece of software which attempts to address these issues by capturing rich

fragments of the WWW onto local storage media. It is possible to put a

collection onto CD ROM, providing portability and inexpensive storage. This

enables the presentation of the WWW to distance learning students, who do not

have internet access. The software interfaces to standard, commonly available

web browsers, acting as a proxy server to the files stored on the local media,

and provides a search engine giving full text searching capability within the

collection.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/12&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Using model

trees for classification&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span class=SpellE&gt;&lt;span

lang=EN-GB&gt;Eibe&lt;/span&gt;&lt;/span&gt;&lt;span lang=EN-GB&gt; Frank, Yong Wang, Stuart &lt;span

class=SpellE&gt;Inglis&lt;/span&gt;, Geoffrey Holmes, Ian H. &lt;span class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Model trees,

which are a type of decision tree with linear regression functions at the

leaves, form the basis of a recent successful technique for predicting

continuous numeric values.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;They can be

applied to classification problems by employing a standard method of

transforming a classification problem into a problem of function

approximation.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Surprisingly, using this

simple transformation the model tree &lt;span class=SpellE&gt;inducer&lt;/span&gt; M5',

based on &lt;span class=SpellE&gt;Quinlan's&lt;/span&gt; M5, generates more accurate

classifiers than the state-of-the-art decision tree learner C5.0, particularly

when most of the attributes are numeric.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/13&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Discovering inter-attribute

relationships&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Geoffrey Holmes&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;It is important

to discover relationships between attributes being used to predict a class

attribute in supervised learning situations for two reasons.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;First, any such relationship will be

potentially interesting to the provider of a &lt;span class=SpellE&gt;dataset&lt;/span&gt;

in its own right.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Second, it would

simplify a learning algorithm's search space, and the related irrelevant

feature and subset selection problem, if the relationships were removed from &lt;span

class=SpellE&gt;datasets&lt;/span&gt; ahead of learning.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;An algorithm to discover such relationships is presented in this

paper.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The algorithm is described and a

surprising number of inter-attribute relationships are discovered in &lt;span

class=SpellE&gt;datasets&lt;/span&gt; from the University of California at Irvine (UCI)

repository.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/14&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Learning from &lt;span

class=SpellE&gt;batched&lt;/span&gt; data:&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;model

combination &lt;span class=SpellE&gt;vs&lt;/span&gt; data combination&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Kai &lt;span

class=SpellE&gt;Ming&lt;/span&gt; Ting, Boon &lt;span class=SpellE&gt;Toh&lt;/span&gt; Low, Ian H. &lt;span

class=SpellE&gt;Witten&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;When presented

with multiple batches of data, one can either combine them into a single batch

before applying a machine learning procedure or learn from each batch

independently and combine the resulting models.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;The former procedure, data combination, is straightforward; this paper

investigates the latter, model combination.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Given an appropriate combination method, one might expect model

combination to prove superior when the data in each batch was obtained under

somewhat different conditions or when different learning algorithms were used

on the batches.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Empirical results show

that model combination often outperforms data combination even when the batches

are drawn randomly from a single source of data and the same learning method is

used on each.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Moreover, this is not just

an &lt;span class=SpellE&gt;artifact&lt;/span&gt; of one particular method of combining

models: it occurs with several different combination methods.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We relate this

phenomenon to the learning curve of the classifiers being used.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Early in the learning process when the

learning curve is steep there is much to gain from data combination, but later

when it becomes shallow there is less to gain and model combination achieves a

greater reduction in variance and hence a lower error rate.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The practical

implication of these results is that one should consider using model

combination rather than data combination, especially when multiple batches of

data for the same task are readily available.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;It is often superior even when the batches are drawn randomly from a

single sample, and we expect its advantage to increase if genuine statistical

differences between the batches exist.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/15&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Information

seeking retrieval, reading and storing behaviour of library users&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Turner K.&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;In the interest

of digital libraries, it is advisable that designers be aware of the potential

behaviour of the users of such a system.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;There are two distinct parts under investigation, the interaction

between traditional libraries involving the seeking and retrieval of relevant

material, and the reading and storage behaviours ensuing. Through this

analysis, the findings could be incorporated into digital library facilities.

There has been copious amounts of research on information seeking leading to

the development of behavioural models to describe the process. Often research

on the information seeking practices of individuals is based on the task and

field of study. The information seeking model, presented by Ellis et al.

(1993), characterises the format of this study where it is used to compare

various research on the information seeking practices of groups of people (from

academics to professionals). It is found that, although researchers do make use

of library facilities, they tend to rely heavily on their own collections and

primarily use the library as a source for previously identified information,

browsing and &lt;span class=SpellE&gt;interloan&lt;/span&gt;. It was found that there are

significant differences in user behaviour between the groups analysed. When

looking at the reading and storage of material it was hard to draw conclusions,

due to the lack of substantial research and information on the topic. However,

through the use of reading strategies, a general idea on how readers behave can

be developed. Designers of digital libraries can benefit from the guidelines

presented here to better understand their audience.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/16&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Proceeding of

the INTERACT97 Combined Workshop on CSCW in HCI-&lt;span class=SpellE&gt;Worldwide&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Matthias &lt;span

class=SpellE&gt;Rauterberg&lt;/span&gt;, Lars &lt;span class=SpellE&gt;Oestreicher&lt;/span&gt;,

John &lt;span class=SpellE&gt;Grundy&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This is the

proceedings for the INTERACT97 combined workshop on “CSCW in HCI-&lt;span

class=SpellE&gt;worldwide&lt;/span&gt;”.&lt;span style='mso-spacerun:yes'&gt;  &lt;/span&gt;The

position papers in this proceedings are those selected from topics relating to

HCI community development &lt;span class=SpellE&gt;worldwide&lt;/span&gt; and to CSCW

issues.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Originally these were to be two

separate INTERACT workshops, but were combined to ensure sufficient

participation for a combined workshop to run.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The combined

workshop has been split into two separate sessions to run in the morning of

July 15&lt;sup&gt;th&lt;/sup&gt;, Sydney, Australia.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;One to discuss the issues relating to the position papers focusing on

general CSCW systems, the other to the development of HCI communities in a &lt;span

class=SpellE&gt;worldwide&lt;/span&gt; context.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;The CSCW session uses as a case study a proposed &lt;span class=SpellE&gt;groupware&lt;/span&gt;

tool for facilitating the development of an HCI database with a &lt;span

class=SpellE&gt;worldwide&lt;/span&gt; geographical distribution.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The HCI community session focuses on

developing the content for such a database, in order for it to foster the

continued development of HCI communities.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;The afternoon session of the combined workshop involves a joint

discussion of the case study &lt;span class=SpellE&gt;groupware&lt;/span&gt; tool, in terms

of its content and likely &lt;span class=SpellE&gt;groupware&lt;/span&gt; facilities.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The position

papers have been grouped into those focusing on HCI communities and hence

content issues for a &lt;span class=SpellE&gt;groupware&lt;/span&gt; database, and those focusing

on CSCW and &lt;span class=SpellE&gt;groupware&lt;/span&gt; issues, and hence likely &lt;span

class=SpellE&gt;groupware&lt;/span&gt; support in the proposed HCI

database/collaboration tools.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We hope

that you find the position papers in this proceedings offer a wide range of

interesting reports of HCI community development &lt;span class=SpellE&gt;worldwide&lt;/span&gt;,

leading CSCW system research, and that a &lt;span class=SpellE&gt;groupware&lt;/span&gt;

tool supporting aspects of a &lt;span class=SpellE&gt;worldwide&lt;/span&gt; HCI database

can draw upon the varied work reported.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/17&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Internationalising

a spreadsheet for Pacific Basin languages&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Robert &lt;span

class=SpellE&gt;Barbour&lt;/span&gt;, Alvin &lt;span class=SpellE&gt;Yeo&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;As people trade

and engage in commerce, an economically dominant culture tends to migrate

language into other recently contacted cultures.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Information technology (IT) can accelerate &lt;span

class=SpellE&gt;enculturation&lt;/span&gt; and promote the expansion of western hegemony

in IT.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Equally, IT can present a

culturally appropriate interface to the user that promotes the preservation of

culture and language with very little additional effort.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In this paper a spreadsheet is

internationalised to accept languages from the Latin-1 character set such as

English, &lt;span class=SpellE&gt;Maori&lt;/span&gt; and &lt;span class=SpellE&gt;Bahasa&lt;/span&gt; &lt;span

class=SpellE&gt;Melayu&lt;/span&gt; (Malaysia’s national language).&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;A technique that allows a non-programmer to

add a new language to the spreadsheet is described.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The technique could also be used to

internationalise other software at the point of design by following the steps

we outline.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/18&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Localising a

spreadsheet:&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;an &lt;span class=SpellE&gt;Iban&lt;/span&gt;

example&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Alvin &lt;span

class=SpellE&gt;Yeo&lt;/span&gt;, Robert &lt;span class=SpellE&gt;Barbour&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Presently, there

is little localisation of software to smaller cultures if it is not

economically viable.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We believe software

should also be localised to the languages of small cultures in order to sustain

and preserve these small cultures.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;As an

example, we localised a spreadsheet from English to &lt;span class=SpellE&gt;Iban&lt;/span&gt;.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The process in which we carried out the

localisation can be used as a framework for the localisation of software to

languages of small ethnic minorities.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Some problems faced during the localisation process are also discussed.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/19&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Strategies of

internationalisation and localisation: a postmodernist/s perspective&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Alvin &lt;span

class=SpellE&gt;Yeo&lt;/span&gt;, Robert &lt;span class=SpellE&gt;Barbour&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Many software

companies today are developing software not only for local consumption but for

the rest of the world.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We introduce the

concepts of internationalisation and localisation and discuss some techniques

using these processes.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;An examination of

&lt;span class=SpellE&gt;postmodern&lt;/span&gt; critique with respect to the software

industry is also reported.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;In addition,

we also feature our proposed internationalisation technique that was inspired

by taking into account the researches of &lt;span class=SpellE&gt;postmodern&lt;/span&gt;

philosophers and mathematicians.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;As illustrated

in our prototype, the technique empowers non-programmers to localise their own

software.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Further development of the

technique and its implications on user interfaces and the future of software

internationalisation and localisation are discussed.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/20&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Language use in

software&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Alvin &lt;span

class=SpellE&gt;Yeo&lt;/span&gt;, Robert &lt;span class=SpellE&gt;Barbour&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Many of the

popular software we use today are in English.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Very few software applications are available in minority languages.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Besides economic goals, we justify why

software should be made available to smaller cultures.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Furthermore, there is evidence that people

learn and progress faster in software in their mother tongue (&lt;span

class=SpellE&gt;Griffiths&lt;/span&gt; et at, 1994) (&lt;span class=SpellE&gt;Krock&lt;/span&gt;,

1996).&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We hypothesise that experienced

users of English spreadsheet can easily migrate to a spreadsheet in their

native tongue i.e. &lt;span class=SpellE&gt;Bahasa&lt;/span&gt; &lt;span class=SpellE&gt;Melayu&lt;/span&gt;

(Malaysia’s national language).&lt;span style='mso-spacerun:yes'&gt; 

&lt;/span&gt;Observations made in the study suggest that the native speakers of &lt;span

class=SpellE&gt;Bahasa&lt;/span&gt; &lt;span class=SpellE&gt;Melayu&lt;/span&gt; had difficulties

with the &lt;span class=SpellE&gt;Bahasa&lt;/span&gt; &lt;span class=SpellE&gt;Melayu&lt;/span&gt;

interface.&lt;span style='mso-spacerun:yes'&gt;  &lt;/span&gt;The subjects’ main difficulty

was their unfamiliarity with computing terminology in &lt;span class=SpellE&gt;Bahasa&lt;/span&gt;

&lt;span class=SpellE&gt;Melayu&lt;/span&gt;.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We

present possible strategies to increase the use of &lt;span class=SpellE&gt;Bahasa&lt;/span&gt;

&lt;span class=SpellE&gt;Melayu&lt;/span&gt; in IT.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;These strategies may also be used to promote the use of other minority

languages in IT.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/21&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Usability

testing:&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;a Malaysian study&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Alvin &lt;span

class=SpellE&gt;Yeo&lt;/span&gt;, Robert &lt;span class=SpellE&gt;Barbour&lt;/span&gt;, Mark &lt;span

class=SpellE&gt;Apperley&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;An exploratory

study of software assessment techniques is conducted in Malaysia.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Subjects in the study comprised staff members

of a Malaysian university with a high Information Technology (IT) presence.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The subjects assessed a spreadsheet tool with

a &lt;span class=SpellE&gt;Bahasa&lt;/span&gt; &lt;span class=SpellE&gt;Melayu&lt;/span&gt; (Malaysia’s

national language) interface.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Software

evaluation techniques used include the think aloud method, interviews and the

System Usability Scale.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The responses in

the various techniques used are reported and initial results indicate

idiosyncratic behaviour of Malaysian subjects.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;The implications of the findings are also discussed.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/22&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Inducing

cost-sensitive trees via instance-weighting&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Kai &lt;span

class=SpellE&gt;Ming&lt;/span&gt; Ting&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;We introduce an

instance-weighting method to induce cost-sensitive trees in this paper.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It is a &lt;span class=SpellE&gt;generalization&lt;/span&gt;

of the standard tree induction process where only the initial instance weights

determine the type of tree (i.e., minimum error trees or minimum cost trees) to

be induced.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We demonstrate that it can

be easily adopted to an existing tree learning algorithm.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Previous

research gave insufficient evidence to support the fact that the greedy

divide-and-conquer algorithm can effectively induce a truly cost-sensitive tree

directly from the training data.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We

provide this empirical evidence in this paper.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;The algorithm employing the instance-weighting method is found to be

comparable to or better than both C4.5 and C5 in terms of total

misclassification costs, tree size and the number of high cost errors.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The instance-weighting method is also simpler

and more effective in implementation than a method based on altered priors.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/23&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Fast convergence

with a greedy tag-phrase dictionary&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Ross &lt;span

class=SpellE&gt;Peeters&lt;/span&gt;, Tony C. Smith&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoBodyText&gt;&lt;span lang=EN-US&gt;The best general-purpose compression

schemes make their gains by estimating a probability distribution over all

possible next symbols given the context established by some number of previous

symbols.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Such context models typically

obtain good compression results for plain text by taking advantage of

regularities in character sequences.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Frequent words and syllables can be incorporated into the model quickly

and thereafter used for reasonably accurate prediction.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;However, the precise context in which

frequent patterns emerge is often extremely varied, and each new word or phrase

immediately introduces new contexts which can adversely affect the compression

rate&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;A great deal of

the structural regularity in a natural language is given rather more by

properties of its grammar than by the orthographic transcription of its

phonology.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;This implies that access to a

grammatical abstraction might lead to good compression.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;While grammatical models have been used

successfully for compressing computer programs [4], grammar-based compression

of plain text has received little attention, primarily because of the

difficulties associated with constructing a suitable natural language

grammar.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;But even without a precise

formulation of the syntax of a language, there is a linguistic abstraction

which is easily accessed and which demonstrates a high degree of regularity

which can be exploited for compression purposes-namely, lexical categories.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/24&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Tag based models

of English text&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;W. J. &lt;span

class=SpellE&gt;Teahan&lt;/span&gt;, John G. &lt;span class=SpellE&gt;Cleary&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The problem of

compressing English text is important both because of the ubiquity of English

as a target for compression and because of the light that compression can shed

on the structure of English.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;English

text is examined in conjunction with additional information about the parts of

speech of each word in the text (these are referred to as “tags”).&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It is shown that the tags plus the text can

be compressed more than the text alone.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;Essentially the tags can be compressed for nothing or even a small net

saving in size.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;A comparison is made of

a number of different ways of integrating compression of tags and text using an

escape mechanism similar to PPM.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;These

are also compared with standard word based and character based compression

programs.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The result is that the tag

character and word based schemes always outperform the character based

schemes.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Overall, the tag based schemes

outperform the word based schemes.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We

conclude by conjecturing that tags chosen for compression rather than

linguistic purposes would perform even better.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/25&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Musical image

compression&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;David

Bainbridge, Stuart &lt;span class=SpellE&gt;Inglis&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Optical music

recognition aims to convert the vast repositories of sheet music in the world

into an on-line digital format [Bai97].&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;In the near future it will be possible to assimilate music into digital

libraries and users will be able to perform searches based on a sung melody in

addition to typical text-based searching [MSW+96].&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;An important requirement for such a system is

the ability to reproduce the original score as accurately as possible.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Due to the huge amount of sheet music

available, the efficient storage of musical images is an important topic of

study.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

investigates whether the “knowledge” extracted from the optical music

recognition (OMR) process can be exploited to gain higher compression than the

JBIG international standard for &lt;span class=SpellE&gt;bi&lt;/span&gt;-level image

compression.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We present a hybrid

approach where the primitive shapes of music extracted by the optical music

recognition process-note heads, note stems, staff lines and so forth-are fed

into a graphical symbol based compression scheme originally designed for images

containing mainly printed text.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Using

this hybrid approach the average compression rate for a single page is improved

by 3.5% over JBIG.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;When multiple pages with

similar typography are processed in sequence, the file size is decreased by

4-8%.&lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Section 2

presents the relevant background to both optical music recognition and textual

image compression.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Section 3 describes

the experiments performed on 66 test images, outlining the combinations of

parameters that were examined to give the best results.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The initial results and refinements are

presented in Section 4, and we conclude in the last section by &lt;span

class=SpellE&gt;summarizing&lt;/span&gt; the findings of this work.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/26&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Correcting English

text using PPM models&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;W. J. &lt;span

class=SpellE&gt;Teahan&lt;/span&gt;, S. &lt;span class=SpellE&gt;Inglis&lt;/span&gt;, J. G. &lt;span

class=SpellE&gt;Cleary&lt;/span&gt;, G. Holmes&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;An essential

component of many applications in natural language processing is a language &lt;span

class=SpellE&gt;modeler&lt;/span&gt; able to correct errors in the text being

processed.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;For optical character recognition

(OCR), poor scanning quality or extraneous pixels in the image may cause one or

more characters to be mis-&lt;span class=SpellE&gt;recognized&lt;/span&gt;; while for

spelling correction, two characters may be transposed, or a character may be

inadvertently inserted or missed out. &lt;/span&gt;&lt;/p&gt;







&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;This paper

describes a method for correcting English text using a PPM model.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;A method that segments words in English text

is introduced and is shown to be a significant improvement over previously used

methods.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;A similar technique is also

applied as a post-processing stage after pages have been &lt;span class=SpellE&gt;recognized&lt;/span&gt;

by a state-of-the-art commercial OCR system.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;We show that the accuracy of the OCR system can be increased from 95.9%

to 96.6%, a decrease of about 10 errors per page.&lt;/span&gt;&lt;/p&gt;



















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/27&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Constraints on

parallelism beyond 10 instructions per cycle&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;John G. &lt;span

class=SpellE&gt;Cleary&lt;/span&gt;, Richard H. &lt;span class=SpellE&gt;Littin&lt;/span&gt;, J. A.

David &lt;span class=SpellE&gt;McWha&lt;/span&gt;, Murray W. Pearson&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The problem of

extracting Instruction Level Parallelism at levels of 10 instructions per clock

and higher is considered.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Two different

architectures which use speculation on memory accesses to achieve this level of

performance are reviewed.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It is pointed

out that while this form of speculation gives high potential parallelism it is

necessary to retain execution state so that incorrect speculation can be detected

and subsequently squashed.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Simulation

results show that the space to store such state is a critical resource in

obtaining good speedup.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;To make good use

of the space it is essential that state be stored efficiently and that it be

retired as soon as possible.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;A number of

techniques for extracting the best usage from the available state storage are

introduced.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/28&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Effects of

re-ordered memory operations on parallelism&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;Richard H. &lt;span

class=SpellE&gt;Littin&lt;/span&gt;, John G. &lt;span class=SpellE&gt;Cleary&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;The performance

effect of permitting different memory operations to be re-ordered is

examined.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The available parallelism is

computed using a machine code simulator.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;A range of possible restrictions on the re-ordering of memory operations

is considered: from the purely sequential case where no re-ordering is

permitted; to the completely permissive one where memory operations may occur

in any order so that the parallelism is restricted only by data

dependencies.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;A general conclusion is

drawn that to reliably obtain parallelism beyond 10 instructions per clock will

require an ability to re-order all memory instructions.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;A brief description of a feasible

architecture capable of this is given.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal style='margin-right:-.4pt'&gt;&lt;span lang=EN-GB&gt;97/29&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;OZCHI’96 Industry Session:&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Sixth Australian Conference on Human-Computer

Interaction&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;Edited by Chris Phillips, Janis &lt;span

class=SpellE&gt;McKauge&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;The idea for a specific industry session at

OZCHI was first mooted at the 1995 conference in &lt;span class=SpellE&gt;Wollongong&lt;/span&gt;,

during questions following a session of short papers which happened

(serendipitously) to be presented by people from industry.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;An animated discussion took place, most of

which was about how OZCHI could be made more relevant to people in industry, be

it working as usability consultants, or working within organisations either as

usability professionals or as ‘champions of the cause’.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The discussion raised more questions than

answers, about the format of such as session, about the challenges of

attracting industry participation, and about the best way of publishing the

results.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Although no real solutions were

arrived at, it was enough to place an industry session on the agenda for

OZCHI’96.&lt;/span&gt;&lt;/p&gt;











&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;97/30&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;Adaptive models of English text&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;W. J. &lt;span class=SpellE&gt;Teahan&lt;/span&gt;,

John G. &lt;span class=SpellE&gt;Cleary&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;High quality models of English text with

performance approaching that of humans is important for many applications

including spelling correction, speech recognition, OCR, and encryption.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;A number of different statistical models of

English are compared with each other and with previous estimates from human

subjects.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;It is concluded that the best

current models are word based with part of speech tags.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Given sufficient training text, they are able

to attain performance comparable to humans.&lt;/span&gt;&lt;/p&gt;















&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;97/31&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;A graphical user interface for Boolean

query specification&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;Steve Jones, &lt;span class=SpellE&gt;Shona&lt;/span&gt;

&lt;span class=SpellE&gt;McInnes&lt;/span&gt;&lt;/span&gt;&lt;/p&gt;



&lt;p class=MsoNormal&gt;&lt;span lang=EN-GB&gt;On-line information repositories commonly

provide keyword search facilities via textual query languages based on Boolean

logic.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;However, there is evidence to

suggest that the syntactical demands of such languages can lead to user errors

and adversely affect the time that it takes users to form queries.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;Users also face difficulties because of the

conflict in semantics between AND &lt;span class=SpellE&gt;and&lt;/span&gt; OR when used in

Boolean logic and English language.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We

suggest that graphical query languages, in particular Venn-like diagrams, can

alleviate the problems that users experience when forming Boolean expressions

with textual languages.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We describe &lt;span

class=SpellE&gt;Vquery&lt;/span&gt;, a Venn-diagram based user interface to the New

Zealand Digital Library (NZDL).&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The

design of &lt;span class=SpellE&gt;Vquery&lt;/span&gt; has been partly motivated by

analysis of NZDL usage.&lt;span style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;We found that

few queries contain more than three terms, use of the intersection operator

dominates and that query refinement is common.&lt;span style='mso-spacerun:yes'&gt;Â 

&lt;/span&gt;A study of the utility of Venn diagrams for query specification

indicates that with little or no training users can interpret and form

Venn-like diagrams which accurately correspond to Boolean expressions.&lt;span

style='mso-spacerun:yes'&gt;Â  &lt;/span&gt;The utility of &lt;span class=SpellE&gt;Vquery&lt;/span&gt;

is considered and directions for future work are proposed.&lt;/span&gt;&lt;/p&gt;











&lt;/div&gt;









</Content>
</Section>
</Archive>