1 | package org.greenstone.atea;
|
---|
2 |
|
---|
3 | import java.util.*;
|
---|
4 | import java.io.*;
|
---|
5 |
|
---|
6 | import org.apache.commons.csv.*;
|
---|
7 | import org.apache.log4j.Logger;
|
---|
8 |
|
---|
9 | import org.greenstone.util.SafeProcess;
|
---|
10 |
|
---|
11 | /**
|
---|
12 | * Program to help going through the n number of random sample web page URLs stored in input
|
---|
13 | * csv file, to eyeball whether the full text (stored in mongodb for each) is indeed inMRI
|
---|
14 | * or not. User can enter Y|N|? and ctr-D or ctrl-C to continue working on this later.
|
---|
15 | * The output file is the input filename + .tmp suffix.
|
---|
16 | * When user continues later, the output file from last time must be used as input file.
|
---|
17 | * Any csv records not completed earlier or with ? entered will be presented for input
|
---|
18 | * on (re-)running this program.
|
---|
19 | *
|
---|
20 | * TO COMPILE OR RUN, FIRST DO:
|
---|
21 | * cd maori-lang-detection/apache-opennlp-1.9.1
|
---|
22 | * export OPENNLP_HOME=`pwd`
|
---|
23 | * cd maori-lang-detection/src
|
---|
24 | *
|
---|
25 | * TO COMPILE:
|
---|
26 | * maori-lang-detection/src$
|
---|
27 | * javac -cp ".:../conf:../lib/*" org/greenstone/atea/ManualURLInspection.java
|
---|
28 | *
|
---|
29 | * TO RUN:
|
---|
30 | * maori-lang-detection/src$
|
---|
31 | * java -cp ".:../conf:../lib/*" org/greenstone/atea/ManualURLInspection ../mongodb-data/random260_manualList_globalDomains_whereAPageContainsMRI.txt
|
---|
32 | *
|
---|
33 | */
|
---|
34 | public class ManualURLInspection {
|
---|
35 |
|
---|
36 | static Logger logger = Logger.getLogger(org.greenstone.atea.ManualURLInspection.class.getName());
|
---|
37 |
|
---|
38 | private final MongoDBQueryer mongodbQueryer;
|
---|
39 | private final File outFolder;
|
---|
40 | private final File webPageURLsCSVFile;
|
---|
41 | private final File tmpOutFile;
|
---|
42 |
|
---|
43 | /** csv column numbers */
|
---|
44 | public static final int URL_COLUMN = 0;
|
---|
45 | public static final int COUNTRY_CODE_COLUMN = 1;
|
---|
46 | public static final int IS_REALLY_IN_MRI_COLUMN = 2;
|
---|
47 | public static final int QUALITY_LEVEL_COLUMN = 3;
|
---|
48 | public static final int COUNT_OF_PAGES_IN_MRI_COLUMN = 4; // count as detected by OpenNLP
|
---|
49 | public static final int TOTAL_PAGES_IN_SITE_COLUMN = 5;
|
---|
50 |
|
---|
51 | /** Possible values for the Quality Level column of the csv file */
|
---|
52 | public static final String NAV = "NAV";
|
---|
53 | public static final String LITTLE_TEXT = "LITTLE_TEXT";
|
---|
54 | public static final String MIXED_TEXT = "MIXED_TEXT";
|
---|
55 | public static final String SIGNIFICANTLY_MAORI = "SIGNIFICANTLY_MAORI";
|
---|
56 | public static final String MAORI_PARAGRAPHS = "MAORI_PARAGRAPHS";
|
---|
57 | public static final String WORDS = "WORDS"; // words or titles, not full sentences
|
---|
58 | public static final String OTHER_LANGUAGES = "OTHER_LANGUAGES";
|
---|
59 | public static final String POEMS_OR_SONGS = "POEMS_OR_SONGS";
|
---|
60 | public static final String SINGLE_MRI_SENTENCE = "SINGLE_MRI_SENTENCE"; // TODO: REVIEW
|
---|
61 | public static final String LINK_TEXT = "LINK_TEXT"; // for office positions designations and link text
|
---|
62 |
|
---|
63 |
|
---|
64 | public ManualURLInspection(MongoDBQueryer mongodbQueryer, File csvFile)
|
---|
65 | {
|
---|
66 | this.mongodbQueryer = mongodbQueryer;
|
---|
67 | this.webPageURLsCSVFile = csvFile;
|
---|
68 | this.outFolder = csvFile.getParentFile();
|
---|
69 |
|
---|
70 | String tmpFilename = Utility.getFilePath(webPageURLsCSVFile);
|
---|
71 | this.tmpOutFile = new File(tmpFilename+".tmp");
|
---|
72 | }
|
---|
73 |
|
---|
74 | public String getCSVOutputFilename() {
|
---|
75 | return (tmpOutFile == null) ? "" : Utility.getFilePath(tmpOutFile);
|
---|
76 | }
|
---|
77 |
|
---|
78 | /**
|
---|
79 | * Read .csv input file one line at a time.
|
---|
80 | * For each line,
|
---|
81 | * - if empty line empty, skip it.
|
---|
82 | * - If the 3rd column of line is already filled in with Y|N, write out identical line
|
---|
83 | * into tmp output file.
|
---|
84 | * - If third column contains ? or if 3rd column is empty, run a MongoDBQuery to get
|
---|
85 | * the full text of the page and display it on screen.
|
---|
86 | * Wait for user input.
|
---|
87 | * - If Enter hit or Y input, write out Y in 3rd field of line into tmp file.
|
---|
88 | * - If N or ? entered, write out N/? as 3rd field.
|
---|
89 | *
|
---|
90 | * Loop through input csv until finished or until Ctrl-C or Ctrl-D pressed.
|
---|
91 | * Ctrl-D here means end of all user interaction, signalling user wants
|
---|
92 | * to stop entering data and continue later.
|
---|
93 | *
|
---|
94 | * When finished or Ctrl-D entered or Ctrl C pressed, all data entered must have been written
|
---|
95 | * out. So to avoid losing data on Ctrl-Ck, write out each processed csv record (whether
|
---|
96 | * already complete or whether user entry made it complete) and flush writer.
|
---|
97 | * When program terminates in any manner, print message that the file has been created.
|
---|
98 | */
|
---|
99 | public String processCSV() {
|
---|
100 |
|
---|
101 | final String USER_PROMPT = "Enter isMRI value of Y|N|? for (%d): %s - %s > ";
|
---|
102 | //"Enter isMRI value of Y|N|? for (" + count + "): " + url + " - " + countryCode + " > ";
|
---|
103 |
|
---|
104 | boolean terminate = false;
|
---|
105 | CSVParser parser = null;
|
---|
106 |
|
---|
107 | try {
|
---|
108 | parser = CSVParser.parse(webPageURLsCSVFile, java.nio.charset.Charset.forName("US-ASCII"), CSVFormat.RFC4180);
|
---|
109 | } catch(Exception e) {
|
---|
110 | logger.error("Failed to parse input CSV file " + Utility.getFilePath(webPageURLsCSVFile), e);
|
---|
111 | return "Failed";
|
---|
112 | }
|
---|
113 |
|
---|
114 | try (
|
---|
115 | //BufferedWriter writer = new BufferedWriter(new FileWriter(tmpOutFile));
|
---|
116 | CSVPrinter csvWriter = new CSVPrinter(new FileWriter(tmpOutFile), CSVFormat.DEFAULT.withQuoteMode(QuoteMode.MINIMAL));
|
---|
117 | ) {
|
---|
118 |
|
---|
119 | int recordCount = 0;
|
---|
120 | for (CSVRecord csvRecord : parser) {
|
---|
121 | //if(terminate) condition handled further below
|
---|
122 |
|
---|
123 | //logger.debug("Got record: " + csvRecord.toString());
|
---|
124 |
|
---|
125 | //int recordNo = csvRecord.RecordNumber(); // will count empty lines!
|
---|
126 |
|
---|
127 | //if(csvRecord.size() != 0) {
|
---|
128 | String url = csvRecord.get(URL_COLUMN);
|
---|
129 | if(url.equals("")) { // skip empty lines
|
---|
130 | continue;
|
---|
131 | }
|
---|
132 |
|
---|
133 | recordCount++;
|
---|
134 | String countryCode = csvRecord.get(COUNTRY_CODE_COLUMN);
|
---|
135 | String isReallyInMRI = "";
|
---|
136 | String qualityLevel = null;
|
---|
137 |
|
---|
138 | //String isReallyInMRI = csvRecord.get(IS_REALLY_IN_MRI_COLUMN);
|
---|
139 | //if(!isReallyInMRI.equals("")) {
|
---|
140 | if(csvRecord.isSet(IS_REALLY_IN_MRI_COLUMN)) {
|
---|
141 | isReallyInMRI = csvRecord.get(IS_REALLY_IN_MRI_COLUMN);
|
---|
142 | }
|
---|
143 |
|
---|
144 | if(csvRecord.isSet(QUALITY_LEVEL_COLUMN)) {
|
---|
145 | qualityLevel = csvRecord.get(QUALITY_LEVEL_COLUMN);
|
---|
146 | }
|
---|
147 |
|
---|
148 | if(terminate || (!isReallyInMRI.equals("") && !isReallyInMRI.equals("?"))) {
|
---|
149 | // if(terminate) on Ctrl-D, don't stop processing csv records
|
---|
150 | // Instead, copy remaining records of input csv file into output csv file
|
---|
151 | isReallyInMRI = isReallyInMRI.toUpperCase();
|
---|
152 | if(qualityLevel == null) {
|
---|
153 | csvWriter.printRecord(url, countryCode, isReallyInMRI);
|
---|
154 | } else {
|
---|
155 | csvWriter.printRecord(url, countryCode, isReallyInMRI, qualityLevel);
|
---|
156 | }
|
---|
157 | csvWriter.flush();
|
---|
158 | logger.info("Got record " + recordCount + ": " + url + " - " + countryCode
|
---|
159 | + " - " + isReallyInMRI + " - " + qualityLevel);
|
---|
160 | }
|
---|
161 | else {
|
---|
162 |
|
---|
163 | // First, display full text for web page record with matching url
|
---|
164 | // so the user can look at it to decide whether it is indeed overall in MRI or not.
|
---|
165 | String fulltext = mongodbQueryer.displayFullTextOfPage(url);
|
---|
166 | System.err.println(String.format("FULL-TEXT for record %d:\n%s\n", recordCount, fulltext));
|
---|
167 |
|
---|
168 | //logger.info("Got record " + recordCount + ": " + url + " - " + countryCode + " - " + qualityLevel);
|
---|
169 |
|
---|
170 | // Read Input until Ctrl-D: read System.In as bufferedReader
|
---|
171 | // https://stackoverflow.com/questions/5837823/read-input-until-controld
|
---|
172 | // Ctrl-C is already taken care if, see
|
---|
173 | // https://coderanch.com/t/279136/java/terminated-program-Control-close-open
|
---|
174 | // "Whenever a process is terminated/killed(CTRL-C), the file descriptors are released. You really do not need to close the stream in such cases."
|
---|
175 | // So I just need to flush the csv print writer after every record is written
|
---|
176 | // and Ctrl-C won't lose any of the data thus far entered by the user.
|
---|
177 |
|
---|
178 | BufferedReader systemIn = new BufferedReader(new InputStreamReader(System.in, "UTF-8"));
|
---|
179 |
|
---|
180 | boolean done = false;
|
---|
181 |
|
---|
182 | System.out.println(String.format(USER_PROMPT, recordCount, url, countryCode));
|
---|
183 | boolean previouslyQuestionMark = false;
|
---|
184 | if(isReallyInMRI.equals("?")) {
|
---|
185 | previouslyQuestionMark = true;
|
---|
186 | System.err.println("\t? entered last time");
|
---|
187 | }
|
---|
188 | while(!done && ((isReallyInMRI = systemIn.readLine()) != null)) {
|
---|
189 |
|
---|
190 | isReallyInMRI = isReallyInMRI.toUpperCase();
|
---|
191 |
|
---|
192 | //logger.debug("@@ Got: |" + isReallyInMRI + "|");
|
---|
193 |
|
---|
194 | // if user hit enter, it means they accepted
|
---|
195 | // - that the full text displayed is really in MRI: Y
|
---|
196 | // - the previous value entered if it was a ?
|
---|
197 | if(isReallyInMRI.equals("")) {
|
---|
198 | if(previouslyQuestionMark) {
|
---|
199 | isReallyInMRI = "?";
|
---|
200 | } else {
|
---|
201 | isReallyInMRI = "Y";
|
---|
202 | }
|
---|
203 | }
|
---|
204 |
|
---|
205 |
|
---|
206 | if(isReallyInMRI.equals("Y") || isReallyInMRI.equals("N") || isReallyInMRI.equals("?")) {
|
---|
207 | done = true;
|
---|
208 | //break;
|
---|
209 | } else {
|
---|
210 | System.out.println("@@ UNRECOGNISED. "
|
---|
211 | + String.format(USER_PROMPT, recordCount, url, countryCode));
|
---|
212 | }
|
---|
213 | }
|
---|
214 |
|
---|
215 | // Save the CSV record - even if quality level is null
|
---|
216 | // Because we don't want to lose the line that used to exist in the file
|
---|
217 | if(qualityLevel == null) {
|
---|
218 | csvWriter.printRecord(url, countryCode, isReallyInMRI);
|
---|
219 | } else {
|
---|
220 | csvWriter.printRecord(url, countryCode, isReallyInMRI, qualityLevel);
|
---|
221 | }
|
---|
222 | csvWriter.flush();
|
---|
223 |
|
---|
224 | if(isReallyInMRI == null) { // if sys.in readLine() was terminated with Ctrl-D
|
---|
225 | terminate = true;
|
---|
226 | System.out.println("User entered Ctrl-D (Lin)/Ctrl-Z (Win) - terminating.");
|
---|
227 | } else {
|
---|
228 | System.out.println("User entered: " + isReallyInMRI);
|
---|
229 | }
|
---|
230 |
|
---|
231 | }
|
---|
232 | }
|
---|
233 | //}
|
---|
234 |
|
---|
235 | } catch(Exception e) {
|
---|
236 | e.printStackTrace();
|
---|
237 | logger.error("Exception occurred when processing CSV file or writing out file:\n"
|
---|
238 | + Utility.getFilePath(tmpOutFile));
|
---|
239 | logger.error(e.getMessage(), e);
|
---|
240 | }
|
---|
241 |
|
---|
242 | //return urlsList;
|
---|
243 | return Utility.getFilePath(tmpOutFile);
|
---|
244 | }
|
---|
245 |
|
---|
246 | /**
|
---|
247 | * Similar to processCSV() above, but for entering the page quality level of each web page
|
---|
248 | * This goes into the QUALITY_LEVEL_COLUMN column of the csv file.
|
---|
249 | * Web pages from some web sites commonly recurring in the csv input file tend to be largely
|
---|
250 | * navigation menus, so preset to NAV. Others are known to be low quality for text resources
|
---|
251 | * as they only have nav menus and pictures despite these being largely in MÄori,
|
---|
252 | * which can also go under NAV.
|
---|
253 | * Other web sites have little text overall whether MÄori or mixed with English, nav included,
|
---|
254 | * (LITTLE_TEXT), or significantly mixed (MRI+ENG/...) text even if a decent amount of text
|
---|
255 | * (MIXED_TEXT). Some sites may largely have standalone words for learning (WORDS).
|
---|
256 | * Other than known websites that have regular content of one of the above types,
|
---|
257 | * the user can enter these values for rarer websites whose web pages may pop up:
|
---|
258 | * NAV, LITTLE_TEXT, MIXED_TEXT, WORDS, SIGNIFICANTLY_MAORI (for decent amounts of MRI text)
|
---|
259 | * MAORI_PARAGRAPHS (for largely continuous paras in MRI even if there are paras in other
|
---|
260 | * langs) and OTHER_LANGUAGES if text not in MRI but mostly in other language,
|
---|
261 | * POEMS_OR_SONGS for content that's largely songs or poetry.
|
---|
262 | */
|
---|
263 | public String processCSV_QualityLevelColumn() {
|
---|
264 |
|
---|
265 | Map<String, String> predefinedDefaultsMap = new HashMap<String, String>();
|
---|
266 | predefinedDefaultsMap.put("tetaurawhiri.govt.nz", NAV);
|
---|
267 | predefinedDefaultsMap.put("tmoa.tki.org.nz", SIGNIFICANTLY_MAORI);
|
---|
268 | predefinedDefaultsMap.put("paekupu.co.nz", MIXED_TEXT); // html is mixed, but display is more MRI
|
---|
269 | predefinedDefaultsMap.put("m.biblepub.com", SIGNIFICANTLY_MAORI);
|
---|
270 | predefinedDefaultsMap.put("biblehub.com", SIGNIFICANTLY_MAORI);
|
---|
271 | predefinedDefaultsMap.put("pukoro.co.nz", WORDS);
|
---|
272 | predefinedDefaultsMap.put("mi.wikipedia.org", MIXED_TEXT);
|
---|
273 | predefinedDefaultsMap.put("mi.m.wikipedia.org", WORDS);
|
---|
274 | predefinedDefaultsMap.put("tkkmmokopuna.school.nz", NAV);
|
---|
275 | predefinedDefaultsMap.put("twtop.school.nz", NAV);
|
---|
276 | predefinedDefaultsMap.put("animations.tewhanake.maori.nz", MAORI_PARAGRAPHS);
|
---|
277 | predefinedDefaultsMap.put("csunplugged.org", SIGNIFICANTLY_MAORI);
|
---|
278 | predefinedDefaultsMap.put("waiata.maori.nz", POEMS_OR_SONGS);
|
---|
279 |
|
---|
280 | final String USER_PROMPT = "Enter qualityLevel value of\n\t? | (N)AV | (L)ITTLE_TEXT | (M)IXED_TEXT | (S)IGNIFICANTLY_MAORI | MAORI_(P)ARAGRAPHS"
|
---|
281 | + "\n\t | LINK_(T)EXT | PO(E)MS_OR_SONGS | S(I)NGLE_MRI_SENTENCE | (W)ORDS | (O)THER_LANGUAGES\n\tfor (%d): %s - %s > ";
|
---|
282 | //"Enter isMRI value of Y|N|? for (" + count + "): " + url + " - " + countryCode + " > ";
|
---|
283 |
|
---|
284 | boolean terminate = false;
|
---|
285 | CSVParser parser = null;
|
---|
286 |
|
---|
287 | try {
|
---|
288 | parser = CSVParser.parse(webPageURLsCSVFile, java.nio.charset.Charset.forName("US-ASCII"), CSVFormat.RFC4180);
|
---|
289 | } catch(Exception e) {
|
---|
290 | logger.error("Failed to parse input CSV file " + Utility.getFilePath(webPageURLsCSVFile), e);
|
---|
291 | return "Failed";
|
---|
292 | }
|
---|
293 |
|
---|
294 | try (
|
---|
295 | CSVPrinter csvWriter = new CSVPrinter(new FileWriter(tmpOutFile), CSVFormat.DEFAULT.withQuoteMode(QuoteMode.MINIMAL));
|
---|
296 | ) {
|
---|
297 |
|
---|
298 | int recordCount = 0;
|
---|
299 | for (CSVRecord csvRecord : parser) {
|
---|
300 | //if(terminate) condition handled further below
|
---|
301 |
|
---|
302 | //logger.debug("Got record: " + csvRecord.toString());
|
---|
303 |
|
---|
304 | String url = csvRecord.get(URL_COLUMN);
|
---|
305 | if(url.equals("")) { // skip empty lines
|
---|
306 | continue;
|
---|
307 | }
|
---|
308 |
|
---|
309 | recordCount++;
|
---|
310 | String countryCode = csvRecord.get(COUNTRY_CODE_COLUMN);
|
---|
311 | String isReallyInMRI = "";
|
---|
312 | String qualityLevel = "";
|
---|
313 |
|
---|
314 | if(csvRecord.isSet(IS_REALLY_IN_MRI_COLUMN)) {
|
---|
315 | isReallyInMRI = csvRecord.get(IS_REALLY_IN_MRI_COLUMN);
|
---|
316 | }
|
---|
317 |
|
---|
318 | if(csvRecord.isSet(QUALITY_LEVEL_COLUMN)) {
|
---|
319 | qualityLevel = csvRecord.get(QUALITY_LEVEL_COLUMN);
|
---|
320 |
|
---|
321 | // Force valid values or ""
|
---|
322 | qualityLevel = getFullQualityLevelNameUppercased(qualityLevel);
|
---|
323 | }
|
---|
324 |
|
---|
325 | if(terminate || (!qualityLevel.equals("") && !qualityLevel.equals("?"))) {
|
---|
326 | // if(terminate) on Ctrl-D, don't stop processing csv records
|
---|
327 | // Instead, copy remaining records of input csv file into output csv file
|
---|
328 |
|
---|
329 | csvWriter.printRecord(url, countryCode, isReallyInMRI, qualityLevel);
|
---|
330 | csvWriter.flush();
|
---|
331 | logger.info("Got record " + recordCount + ": " + url + " - " + countryCode
|
---|
332 | + " - " + isReallyInMRI + " - " + qualityLevel);
|
---|
333 | }
|
---|
334 | else {
|
---|
335 |
|
---|
336 | // First, display full text for web page record with matching url
|
---|
337 | // so the user can look at it to decide whether it is indeed overall in MRI or not.
|
---|
338 | String fulltext = mongodbQueryer.displayFullTextOfPage(url);
|
---|
339 | System.err.println(String.format("\nFULL-TEXT for record %d:\n%s\n", recordCount, fulltext));
|
---|
340 |
|
---|
341 | //logger.info("Got record " + recordCount + ": " + url + " - " + countryCode + " - " + qualityLevel);
|
---|
342 |
|
---|
343 | // Read Input until Ctrl-D: read System.In as bufferedReader
|
---|
344 | // https://stackoverflow.com/questions/5837823/read-input-until-controld
|
---|
345 | // Ctrl-C is already taken care if, see
|
---|
346 | // https://coderanch.com/t/279136/java/terminated-program-Control-close-open
|
---|
347 | // "Whenever a process is terminated/killed(CTRL-C), the file descriptors are released. You really do not need to close the stream in such cases."
|
---|
348 | // So I just need to flush the csv print writer after every record is written
|
---|
349 | // and Ctrl-C won't lose any of the data thus far entered by the user.
|
---|
350 |
|
---|
351 | BufferedReader systemIn = new BufferedReader(new InputStreamReader(System.in, "UTF-8"));
|
---|
352 |
|
---|
353 | boolean done = false;
|
---|
354 |
|
---|
355 | // Work out default if basic URLs present in defaults map
|
---|
356 | // If it is, use its value as default for this URL
|
---|
357 | String basicURL = Utility.stripProtocolAndWWWFromURL(Utility.getDomainForURL(url, false));
|
---|
358 | String predefQualityLevel = predefinedDefaultsMap.get(basicURL);
|
---|
359 |
|
---|
360 | System.out.println(String.format(USER_PROMPT, recordCount, url, countryCode));
|
---|
361 | if(predefQualityLevel != null) {
|
---|
362 | System.err.println("\tDefault for this domain: " + predefQualityLevel
|
---|
363 | + ". Press Enter to accept >");
|
---|
364 | }
|
---|
365 |
|
---|
366 | boolean previouslyQuestionMark = false;
|
---|
367 | String oldQualityLevel = qualityLevel;
|
---|
368 |
|
---|
369 | if(qualityLevel.equals("?")) {
|
---|
370 | previouslyQuestionMark = true;
|
---|
371 | System.err.println("\t? entered last time. Press Enter to keep >");
|
---|
372 | }
|
---|
373 | while(!done && ((qualityLevel = systemIn.readLine()) != null)) {
|
---|
374 | //logger.debug("@@ Got: |" + qualityLevel + "|");
|
---|
375 |
|
---|
376 | // If the user hit enter, it means they accepted
|
---|
377 | // - the previous value entered, if it was a ?
|
---|
378 | // - or want the default for the URL if any displayed
|
---|
379 | // - or want SIGNIFICANTLY_MAORI if no default displayed
|
---|
380 | if(qualityLevel.equals("")) { // User just hit enter without other chars
|
---|
381 | if(previouslyQuestionMark) {
|
---|
382 | qualityLevel = "?";
|
---|
383 | } else {
|
---|
384 | qualityLevel = (predefQualityLevel == null) ? SIGNIFICANTLY_MAORI : predefQualityLevel;
|
---|
385 | }
|
---|
386 |
|
---|
387 | oldQualityLevel = qualityLevel;
|
---|
388 | }
|
---|
389 | else {
|
---|
390 | // force valid values - will return "" if invalid value
|
---|
391 | qualityLevel = getFullQualityLevelNameUppercased(qualityLevel);
|
---|
392 | }
|
---|
393 |
|
---|
394 | // only if qualityLevel entered was invalid, would it now
|
---|
395 | // have been changed to ""
|
---|
396 | if(!qualityLevel.equals("")) {
|
---|
397 | oldQualityLevel = qualityLevel;
|
---|
398 | done = true;
|
---|
399 | } else {
|
---|
400 | System.out.println("@@ UNRECOGNISED. "
|
---|
401 | + String.format(USER_PROMPT, recordCount, url, countryCode));
|
---|
402 | }
|
---|
403 | }
|
---|
404 |
|
---|
405 | // Save the CSV record - even if quality level is null
|
---|
406 | // Because we don't want to lose the line that used to exist in the file
|
---|
407 | csvWriter.printRecord(url, countryCode, isReallyInMRI, qualityLevel);
|
---|
408 | csvWriter.flush();
|
---|
409 |
|
---|
410 | if(qualityLevel == null) { // if sys.in readLine() was terminated with Ctrl-D
|
---|
411 | terminate = true;
|
---|
412 | System.out.println("--- Got Ctrl-D (Lin)/Ctrl-Z (Win). Terminating. ---");
|
---|
413 | } else {
|
---|
414 | System.out.println("User entered: " + oldQualityLevel);
|
---|
415 |
|
---|
416 | }
|
---|
417 | }
|
---|
418 | }
|
---|
419 |
|
---|
420 | if(terminate = true) {
|
---|
421 | System.out.println("User entered Ctrl-D (Lin)/Ctrl-Z (Win) - terminating.");
|
---|
422 | }
|
---|
423 |
|
---|
424 | } catch(Exception e) {
|
---|
425 | e.printStackTrace();
|
---|
426 | logger.error("Exception occurred when processing CSV file or writing out file:\n"
|
---|
427 | + Utility.getFilePath(tmpOutFile));
|
---|
428 | logger.error(e.getMessage(), e);
|
---|
429 | }
|
---|
430 |
|
---|
431 |
|
---|
432 | return Utility.getFilePath(tmpOutFile);
|
---|
433 | }
|
---|
434 |
|
---|
435 | public String getFullQualityLevelNameUppercased(String qualityLevel) {
|
---|
436 |
|
---|
437 | qualityLevel = qualityLevel.toUpperCase();
|
---|
438 |
|
---|
439 | if(qualityLevel.equals("N")) {
|
---|
440 | return NAV;
|
---|
441 | } else if(qualityLevel.equals("L")) {
|
---|
442 | return LITTLE_TEXT;
|
---|
443 | } else if(qualityLevel.equals("M")) {
|
---|
444 | return MIXED_TEXT;
|
---|
445 | } else if(qualityLevel.equals("S")) {
|
---|
446 | return SIGNIFICANTLY_MAORI;
|
---|
447 | } else if(qualityLevel.equals("P")) {
|
---|
448 | return MAORI_PARAGRAPHS;
|
---|
449 | } else if(qualityLevel.equals("W")) {
|
---|
450 | return WORDS;
|
---|
451 | } else if(qualityLevel.equals("O")) {
|
---|
452 | return OTHER_LANGUAGES;
|
---|
453 | } else if(qualityLevel.equals("E")) {
|
---|
454 | return POEMS_OR_SONGS;
|
---|
455 | } else if(qualityLevel.equals("I")) {
|
---|
456 | return SINGLE_MRI_SENTENCE;
|
---|
457 | } else if(qualityLevel.equals("T")) {
|
---|
458 | return LINK_TEXT;
|
---|
459 | } else if(qualityLevel.equals(NAV)
|
---|
460 | || qualityLevel.equals(LITTLE_TEXT)
|
---|
461 | || qualityLevel.equals(MIXED_TEXT)
|
---|
462 | || qualityLevel.equals(SIGNIFICANTLY_MAORI)
|
---|
463 | || qualityLevel.equals(MAORI_PARAGRAPHS)
|
---|
464 | || qualityLevel.equals(WORDS)
|
---|
465 | || qualityLevel.equals(OTHER_LANGUAGES)
|
---|
466 | || qualityLevel.equals(POEMS_OR_SONGS)
|
---|
467 | || qualityLevel.equals(SINGLE_MRI_SENTENCE)
|
---|
468 | || qualityLevel.equals(LINK_TEXT)) {
|
---|
469 | return qualityLevel;
|
---|
470 | }
|
---|
471 | return "";
|
---|
472 | }
|
---|
473 |
|
---|
474 |
|
---|
475 | public void reviewQualityLevelFieldFor(/*String basicDomain,*/ String fieldValue) {
|
---|
476 | final String USER_PROMPT = "Enter qualityLevel value of\n\t? | (N)AV | (L)ITTLE_TEXT | (M)IXED_TEXT | (S)IGNIFICANTLY_MAORI | MAORI_(P)ARAGRAPHS"
|
---|
477 | + "\n\t | LINK_(T)EXT | PO(E)MS_OR_SONGS | S(I)NGLE_MRI_SENTENCE | (W)ORDS | (O)THER_LANGUAGES\n\tfor (%d): %s - %s > ";
|
---|
478 | //"Enter isMRI value of Y|N|? for (" + count + "): " + url + " - " + countryCode + " > ";
|
---|
479 |
|
---|
480 | boolean terminate = false;
|
---|
481 | CSVParser parser = null;
|
---|
482 |
|
---|
483 | try {
|
---|
484 | parser = CSVParser.parse(webPageURLsCSVFile, java.nio.charset.Charset.forName("US-ASCII"), CSVFormat.RFC4180);
|
---|
485 | } catch(Exception e) {
|
---|
486 | logger.error("Failed to parse input CSV file " + Utility.getFilePath(webPageURLsCSVFile), e);
|
---|
487 | return;
|
---|
488 | }
|
---|
489 |
|
---|
490 | try (
|
---|
491 | CSVPrinter csvWriter = new CSVPrinter(new FileWriter(tmpOutFile), CSVFormat.DEFAULT.withQuoteMode(QuoteMode.MINIMAL));
|
---|
492 | ) {
|
---|
493 |
|
---|
494 | int recordCount = 0;
|
---|
495 | for (CSVRecord csvRecord : parser) {
|
---|
496 |
|
---|
497 | String url = csvRecord.get(URL_COLUMN);
|
---|
498 | if(url.equals("")) { // skip empty lines
|
---|
499 | continue;
|
---|
500 | }
|
---|
501 |
|
---|
502 | recordCount++;
|
---|
503 |
|
---|
504 | String basicURL = Utility.stripProtocolAndWWWFromURL(Utility.getDomainForURL(url, false));
|
---|
505 | /*
|
---|
506 | if(!basicURL.equals(basicDomain)) {
|
---|
507 | continue; // skip URLs we're not interested in
|
---|
508 | }
|
---|
509 | */
|
---|
510 |
|
---|
511 | // Work out default if basic URLs present in defaults map
|
---|
512 | // If it is, use its value as default for this URL
|
---|
513 | //String predefQualityLevel = predefinedDefaultsMap.get(basicURL);
|
---|
514 |
|
---|
515 | String countryCode = csvRecord.get(COUNTRY_CODE_COLUMN);
|
---|
516 | String isReallyInMRI = "";
|
---|
517 | String qualityLevel = "";
|
---|
518 |
|
---|
519 | if(csvRecord.isSet(IS_REALLY_IN_MRI_COLUMN)) {
|
---|
520 | isReallyInMRI = csvRecord.get(IS_REALLY_IN_MRI_COLUMN);
|
---|
521 | }
|
---|
522 |
|
---|
523 | if(csvRecord.isSet(QUALITY_LEVEL_COLUMN)) {
|
---|
524 | qualityLevel = csvRecord.get(QUALITY_LEVEL_COLUMN);
|
---|
525 |
|
---|
526 | // Force valid values or ""
|
---|
527 | qualityLevel = getFullQualityLevelNameUppercased(qualityLevel);
|
---|
528 | }
|
---|
529 |
|
---|
530 |
|
---|
531 | if(terminate || !qualityLevel.equals(fieldValue)
|
---|
532 | /* || basicURL.equals("paekupu.co.nz") // when reviewing MIXED_TEXT */
|
---|
533 | /*|| basicURL.equals("tetaurawhiri.govt.nz") // when reviewing NAV */
|
---|
534 | /*|| basicURL.equals("biblehub.com") || basicURL.equals("m.biblepub.com") // when reviewing SIGNIFICANTLY_MAORI */) {
|
---|
535 | // if(terminate) on Ctrl-D, don't stop processing csv records
|
---|
536 | // Instead, copy remaining records of input csv file into output csv file
|
---|
537 |
|
---|
538 | // Similarly, if the qualityLevel field does not have the value we're interested in
|
---|
539 | // then just write it out as-is
|
---|
540 | csvWriter.printRecord(url, countryCode, isReallyInMRI, qualityLevel);
|
---|
541 | csvWriter.flush();
|
---|
542 | logger.info("Got record " + recordCount + ": " + url + " - " + countryCode
|
---|
543 | + " - " + isReallyInMRI + " - " + qualityLevel);
|
---|
544 | }
|
---|
545 | else {
|
---|
546 |
|
---|
547 | // First, display full text for web page record with matching url
|
---|
548 | // so the user can look at it to decide whether it is indeed overall in MRI or not.
|
---|
549 | String fulltext = mongodbQueryer.displayFullTextOfPage(url);
|
---|
550 | System.err.println(String.format("\nFULL-TEXT for record %d:\n%s\n", recordCount, fulltext));
|
---|
551 |
|
---|
552 | //logger.info("Got record " + recordCount + ": " + url + " - " + countryCode + " - " + qualityLevel);
|
---|
553 |
|
---|
554 | // Read Input until Ctrl-D: read System.In as bufferedReader
|
---|
555 | // https://stackoverflow.com/questions/5837823/read-input-until-controld
|
---|
556 | // Ctrl-C is already taken care if, see
|
---|
557 | // https://coderanch.com/t/279136/java/terminated-program-Control-close-open
|
---|
558 | // "Whenever a process is terminated/killed(CTRL-C), the file descriptors are released. You really do not need to close the stream in such cases."
|
---|
559 | // So I just need to flush the csv print writer after every record is written
|
---|
560 | // and Ctrl-C won't lose any of the data thus far entered by the user.
|
---|
561 |
|
---|
562 | BufferedReader systemIn = new BufferedReader(new InputStreamReader(System.in, "UTF-8"));
|
---|
563 |
|
---|
564 | boolean done = false;
|
---|
565 |
|
---|
566 | System.out.println(String.format(USER_PROMPT, recordCount, url, countryCode));
|
---|
567 | //if(predefQualityLevel != null) {
|
---|
568 | //System.err.println("\tDefault for this domain: " + predefQualityLevel
|
---|
569 | //+ ". Press Enter to accept >");
|
---|
570 | //}
|
---|
571 |
|
---|
572 | if(qualityLevel.equals(fieldValue)) {
|
---|
573 | System.err.println("\t" + fieldValue + " entered last time. Press Enter to keep >");
|
---|
574 | }
|
---|
575 | while(!done && ((qualityLevel = systemIn.readLine()) != null)) {
|
---|
576 | //logger.debug("@@ Got: |" + qualityLevel + "|");
|
---|
577 |
|
---|
578 | // If the user hit enter, it means they accepted the previous value entered
|
---|
579 | if(qualityLevel.equals("")) { // User just hit enter without other chars
|
---|
580 | qualityLevel = fieldValue;
|
---|
581 | }
|
---|
582 | else {
|
---|
583 | // force valid values - will return "" if invalid value
|
---|
584 | qualityLevel = getFullQualityLevelNameUppercased(qualityLevel);
|
---|
585 | }
|
---|
586 |
|
---|
587 | // only if qualityLevel entered was invalid, would it now
|
---|
588 | // have been changed to ""
|
---|
589 | if(!qualityLevel.equals("")) {
|
---|
590 | done = true;
|
---|
591 | } else {
|
---|
592 | System.out.println("@@ UNRECOGNISED. "
|
---|
593 | + String.format(USER_PROMPT, recordCount, url, countryCode));
|
---|
594 | }
|
---|
595 | }
|
---|
596 |
|
---|
597 | // Save the CSV record - even if quality level is null
|
---|
598 | // Because we don't want to lose the line that used to exist in the file
|
---|
599 | csvWriter.printRecord(url, countryCode, isReallyInMRI, qualityLevel);
|
---|
600 | csvWriter.flush();
|
---|
601 |
|
---|
602 | if(qualityLevel == null) { // if sys.in readLine() was terminated with Ctrl-D
|
---|
603 | terminate = true;
|
---|
604 | System.out.println("--- Got Ctrl-D (Lin)/Ctrl-Z (Win). Terminating. ---");
|
---|
605 | } else {
|
---|
606 | System.out.println("User entered: " + qualityLevel);
|
---|
607 |
|
---|
608 | }
|
---|
609 | }
|
---|
610 | }
|
---|
611 |
|
---|
612 | if(terminate = true) {
|
---|
613 | System.out.println("User entered Ctrl-D (Lin)/Ctrl-Z (Win) - terminating.");
|
---|
614 | }
|
---|
615 |
|
---|
616 | } catch(Exception e) {
|
---|
617 | e.printStackTrace();
|
---|
618 | logger.error("Exception occurred when processing CSV file or writing out file:\n"
|
---|
619 | + Utility.getFilePath(tmpOutFile));
|
---|
620 | logger.error(e.getMessage(), e);
|
---|
621 | }
|
---|
622 |
|
---|
623 | }
|
---|
624 |
|
---|
625 | /**
|
---|
626 | * Add 2 new columns to the csv file: num pages in site that are inMRI and total num pages in site.
|
---|
627 | */
|
---|
628 | public void insertTotalsIntoCSVRecords() {
|
---|
629 |
|
---|
630 | boolean terminate = false;
|
---|
631 | CSVParser parser = null;
|
---|
632 |
|
---|
633 | try {
|
---|
634 | parser = CSVParser.parse(webPageURLsCSVFile, java.nio.charset.Charset.forName("US-ASCII"), CSVFormat.RFC4180);
|
---|
635 | } catch(Exception e) {
|
---|
636 | logger.error("Failed to parse input CSV file " + Utility.getFilePath(webPageURLsCSVFile), e);
|
---|
637 | return;
|
---|
638 | }
|
---|
639 |
|
---|
640 | try (
|
---|
641 | CSVPrinter csvWriter = new CSVPrinter(new FileWriter(tmpOutFile), CSVFormat.DEFAULT.withQuoteMode(QuoteMode.MINIMAL));
|
---|
642 | ) {
|
---|
643 |
|
---|
644 |
|
---|
645 | int recordCount = 0;
|
---|
646 |
|
---|
647 | csvWriter.printRecord("origSequence", "basicDomain","pageURL",
|
---|
648 | "countryCode","mostlyMRI","qualityLevel",
|
---|
649 | "numPagesInMRIForDomainSuffix","totalPagesInMRIForDomainSuffix");
|
---|
650 |
|
---|
651 |
|
---|
652 | for (CSVRecord csvRecord : parser) {
|
---|
653 |
|
---|
654 | String url = csvRecord.get(URL_COLUMN);
|
---|
655 | if(url.equals("")) { // skip empty lines
|
---|
656 | continue;
|
---|
657 | }
|
---|
658 | recordCount++;
|
---|
659 |
|
---|
660 | String basicURL = Utility.stripProtocolAndWWWFromURL(Utility.getDomainForURL(url, false));
|
---|
661 |
|
---|
662 | String countryCode = csvRecord.get(COUNTRY_CODE_COLUMN);
|
---|
663 | String isReallyInMRI = "";
|
---|
664 | String qualityLevel = "";
|
---|
665 |
|
---|
666 | if(csvRecord.isSet(IS_REALLY_IN_MRI_COLUMN)) {
|
---|
667 | isReallyInMRI = csvRecord.get(IS_REALLY_IN_MRI_COLUMN);
|
---|
668 | }
|
---|
669 |
|
---|
670 | if(csvRecord.isSet(QUALITY_LEVEL_COLUMN)) {
|
---|
671 | qualityLevel = csvRecord.get(QUALITY_LEVEL_COLUMN);
|
---|
672 | }
|
---|
673 |
|
---|
674 | //COUNT_OF_PAGES_IN_MRI_COLUMN; TOTAL_PAGES_IN_SITE_COLUMN;
|
---|
675 | long countNumPagesInMRI = mongodbQueryer.getFieldTotalForDomainSuffix(
|
---|
676 | basicURL, MongoDBQueryer.FIELD_NUM_PAGES_IN_MRI);
|
---|
677 | long countTotalPages = mongodbQueryer.getFieldTotalForDomainSuffix(
|
---|
678 | basicURL, MongoDBQueryer.FIELD_TOTAL_PAGES);
|
---|
679 |
|
---|
680 | logger.info("Got record " + recordCount + ": " + url + " - " + countryCode
|
---|
681 | + " - " + isReallyInMRI + " - " + qualityLevel
|
---|
682 | + " - " + countNumPagesInMRI + " - " + countTotalPages);
|
---|
683 |
|
---|
684 | // Save the CSV record into the tmp file with the 2 counts columns
|
---|
685 | csvWriter.printRecord(recordCount, basicURL,
|
---|
686 | url, countryCode, isReallyInMRI, qualityLevel,
|
---|
687 | countNumPagesInMRI, countTotalPages);
|
---|
688 | }
|
---|
689 |
|
---|
690 | } catch(Exception e) {
|
---|
691 | e.printStackTrace();
|
---|
692 | logger.error("Exception occurred when processing CSV file or writing out file:\n"
|
---|
693 | + Utility.getFilePath(tmpOutFile));
|
---|
694 | logger.error(e.getMessage(), e);
|
---|
695 | }
|
---|
696 |
|
---|
697 | }
|
---|
698 |
|
---|
699 |
|
---|
700 | public static void printUsage() {
|
---|
701 | System.err.println("Usage: ManualURLInspection webPageURLs.txt");
|
---|
702 | }
|
---|
703 |
|
---|
704 | //^https?://(www.)?
|
---|
705 | /**
|
---|
706 | * If no args are passed in, generates complete containsMRI file listings for NZ and overseas web SITES (domains),
|
---|
707 | * with overseas web sites that have mi (mi.* or *\/mi) in the URL path listed separately.
|
---|
708 | * You can then manually inspect the domains in this listing to shortlist which of these sites are not automatically
|
---|
709 | * translated and really contain at least one webpage containing at least one sentence in MRI.
|
---|
710 | * If a file is passed in containing a list of domains, then this first generates a full listing of all webpages
|
---|
711 | * matching isMRI for each site in the domain list. It then generates a smaller set of random webpages matching
|
---|
712 | * isMRI for the pooled sites in the domain list where the sample size of URLs produced is sufficient for giving
|
---|
713 | * 90% confidence with 5% margin of error for testing binary outcomes, see
|
---|
714 | * https://stats.stackexchange.com/questions/207584/sample-size-choice-with-binary-outcome
|
---|
715 | */
|
---|
716 | public static void main(String args[]) {
|
---|
717 | SafeProcess.DEBUG = 1;
|
---|
718 |
|
---|
719 | if(args.length != 1) {
|
---|
720 | printUsage();
|
---|
721 | System.exit(-1);
|
---|
722 | }
|
---|
723 |
|
---|
724 |
|
---|
725 | try (
|
---|
726 | MongoDBQueryer mongodb = new MongoDBQueryer();
|
---|
727 | ) {
|
---|
728 |
|
---|
729 | mongodb.connectToDB();
|
---|
730 |
|
---|
731 | // output files will be stored in mongodb-data-auto
|
---|
732 | File outFolder = new File("../mongodb-data-auto/").getAbsoluteFile();
|
---|
733 |
|
---|
734 |
|
---|
735 | logger.info("*************************************");
|
---|
736 |
|
---|
737 |
|
---|
738 |
|
---|
739 | final File inputFile = new File(args[0]);
|
---|
740 | if(!inputFile.exists()) {
|
---|
741 | logger.info("File " + inputFile + " does not exist");
|
---|
742 | System.exit(-1);
|
---|
743 | }
|
---|
744 |
|
---|
745 | final ManualURLInspection inspector = new ManualURLInspection(mongodb, inputFile);
|
---|
746 |
|
---|
747 | Runtime.getRuntime().addShutdownHook(new Thread(new Runnable() {
|
---|
748 | public void run() {
|
---|
749 | logger.info("@@@@@@@@@@@@@@@@@@@@@@@@");
|
---|
750 | logger.info("WARNING: If Ctrl-C was pressed, then");
|
---|
751 | logger.info("\tan INCOMPLETE temp CSV file would have been generated at: " +
|
---|
752 | inspector.getCSVOutputFilename());
|
---|
753 | logger.info(String.format("\tSo copy remaining records from input file %s into this file.",
|
---|
754 | Utility.getFilePath(inputFile)));
|
---|
755 | logger.info("@@@@@@@@@@@@@@@@@@@@@@@@");
|
---|
756 | }
|
---|
757 | }));
|
---|
758 |
|
---|
759 | //String filename = inspector.processCSV();
|
---|
760 | //String filename = inspector.processCSV_QualityLevelColumn();
|
---|
761 |
|
---|
762 |
|
---|
763 | //inspector.reviewQualityLevelFieldFor("SINGLE_MRI_SENTENCE");
|
---|
764 |
|
---|
765 | inspector.insertTotalsIntoCSVRecords();
|
---|
766 |
|
---|
767 | //logger.info("Generated temp CSV file: " + filename);
|
---|
768 | logger.info("*************************************");
|
---|
769 | } catch(Exception e) {
|
---|
770 | logger.error(e.getMessage(), e);
|
---|
771 | }
|
---|
772 | }
|
---|
773 | }
|
---|