1 | package org.greenstone.atea;
|
---|
2 |
|
---|
3 | import java.io.*;
|
---|
4 | import java.lang.ArrayIndexOutOfBoundsException;
|
---|
5 | import java.util.ArrayList;
|
---|
6 | import java.util.Arrays;
|
---|
7 |
|
---|
8 | import org.apache.log4j.Logger;
|
---|
9 |
|
---|
10 | /**
|
---|
11 | * Class to process the dump text files produced FOR EACH SITE (e.g. site "00001") that
|
---|
12 | * Nutch has finished crawling and whose text has been dumped out to a file called dump.txt.
|
---|
13 | * This reads in the dump.txt file contained in each site folder within the input folder.
|
---|
14 | * (e.g. input folder "crawled" could contain folders 00001 to 01465. Each contains a dump.txt)
|
---|
15 | * Each dump.txt could contain the text contents for an entire site, or for individual pages.
|
---|
16 | * This class then uses class TextDumpPage to parse each webpage within a dump.txt,
|
---|
17 | * which parses out the actual text body content of each webpage's section within a dump.txt.
|
---|
18 | * Finally, MaoriTextDetector is run over that to determine whether the full body text is
|
---|
19 | * likely to be in Maori or not.
|
---|
20 | *
|
---|
21 | * Potential issues: since a web page's text is dumped out by nutch with neither paragraph
|
---|
22 | * nor even newline separator, it's hard to be sure that the entire page is in language.
|
---|
23 | * If it's in multiple languages, there's no way to be sure there aren't promising Maori language
|
---|
24 | * paragraphs contained in a page, if the majority/the remainder happen to be in English.
|
---|
25 | *
|
---|
26 | * So if we're looking for any paragraphs in Maori to store in a DB, perhaps it's better to run
|
---|
27 | * the MaoriTextDetector.isTextInMaori(BufferedReader reader) over two "lines" at a time,
|
---|
28 | * instead of running it over the entire html body's text.
|
---|
29 | *
|
---|
30 | * TO COMPILE OR RUN, FIRST DO:
|
---|
31 | * cd maori-lang-detection/apache-opennlp-1.9.1
|
---|
32 | * export OPENNLP_HOME=`pwd`
|
---|
33 | * cd maori-lang-detection/src
|
---|
34 | *
|
---|
35 | * TO COMPILE:
|
---|
36 | * maori-lang-detection/src$
|
---|
37 | * javac -cp ".:../conf:../lib/*:$OPENNLP_HOME/lib/opennlp-tools-1.9.1.jar" org/greenstone/atea/NutchTextDumpProcessor.java
|
---|
38 | *
|
---|
39 | * TO RUN:
|
---|
40 | * maori-lang-detection/src$
|
---|
41 | * java -cp ".:../conf:../lib/*:$OPENNLP_HOME/lib/opennlp-tools-1.9.1.jar" org/greenstone/atea/NutchTextDumpProcessor ../crawled
|
---|
42 | *
|
---|
43 | * or:
|
---|
44 | * java -cp ".:../conf:../lib/*:$OPENNLP_HOME/lib/opennlp-tools-1.9.1.jar" org/greenstone/atea/NutchTextDumpProcessor ../crawled > ../crawled/bla.txt 2>&1
|
---|
45 | *
|
---|
46 | */
|
---|
47 | public class NutchTextDumpProcessor {
|
---|
48 | private static Logger logger = Logger.getLogger(org.greenstone.atea.NutchTextDumpProcessor.class.getName());
|
---|
49 |
|
---|
50 | static boolean DEBUG_MODE = true;
|
---|
51 |
|
---|
52 | private final MaoriTextDetector maoriTxtDetector;
|
---|
53 |
|
---|
54 | public final String siteID; // is this necessary?
|
---|
55 | private String domainOfSite;
|
---|
56 |
|
---|
57 | /** keep a list to store the text of each page */
|
---|
58 | private ArrayList<TextDumpPage> pages;
|
---|
59 |
|
---|
60 | /** list of pages in this site which were detected as being in MRI */
|
---|
61 | private ArrayList<MRIWebPageStats> pagesInMRI;
|
---|
62 | /** list of pages in this site which were NOT detected as being in MRI but nevertheless
|
---|
63 | * contain one or more sentences in MRI
|
---|
64 | */
|
---|
65 | private ArrayList<MRIWebPageStats> pagesContainingMRI;
|
---|
66 |
|
---|
67 | private boolean isStartOfNewWebPageRecord(String prevLine, String line) {
|
---|
68 | // The start of a new web page's record in nutch's text dump of an entire site
|
---|
69 | // is denoted by a newline followed by a URL (protocol)
|
---|
70 | // or the very start of the file with a URL (protocol)
|
---|
71 | return ((prevLine == null || prevLine.equals(""))
|
---|
72 | && (line.startsWith("http://") || line.startsWith("https://")));
|
---|
73 | }
|
---|
74 |
|
---|
75 | public void debugPageDump(StringBuilder pageDump) {
|
---|
76 | if(DEBUG_MODE) {
|
---|
77 | // START DEBUG
|
---|
78 | debug("__________________________________________");
|
---|
79 | debug("@@@ Found page entry: ");
|
---|
80 | debug("__________________________________________");
|
---|
81 | debug(pageDump.toString());
|
---|
82 | debug("------------------------------------------");
|
---|
83 | // END DEBUG
|
---|
84 | }
|
---|
85 | }
|
---|
86 |
|
---|
87 | public NutchTextDumpProcessor(MaoriTextDetector maoriTxtDetector, String siteID, File txtDumpFile) {
|
---|
88 | // siteID is of the form %5d (e.g. 00020) and is just the name of a site folder
|
---|
89 | this.siteID = siteID;
|
---|
90 | this.maoriTxtDetector = maoriTxtDetector;
|
---|
91 |
|
---|
92 | pages = new ArrayList<TextDumpPage>();
|
---|
93 |
|
---|
94 | String line = null;
|
---|
95 | StringBuilder pageDump = null;
|
---|
96 | try (
|
---|
97 | BufferedReader reader = new BufferedReader(new FileReader(txtDumpFile));
|
---|
98 | ) {
|
---|
99 |
|
---|
100 | boolean readingText = false;
|
---|
101 | String prevLine = null;
|
---|
102 |
|
---|
103 | while((line = reader.readLine()) != null) { // readLine removes newline separator
|
---|
104 | line = line.trim();
|
---|
105 | // iff outside of a page's body text, then an empty line marks the end of a page
|
---|
106 | // in nutch's text dump of a site.
|
---|
107 | // But note, there can be an empty line (or more?) between the start and end
|
---|
108 | // markers of a page's text, though.
|
---|
109 |
|
---|
110 | if(isStartOfNewWebPageRecord(prevLine, line)) {
|
---|
111 |
|
---|
112 | if(pageDump != null) { // should also be the case then: if(prevLine != null)
|
---|
113 | // finish old pageDump and begin new one
|
---|
114 |
|
---|
115 | //debugPageDump(pageDump);
|
---|
116 |
|
---|
117 | TextDumpPage page = new TextDumpPage(siteID, pageDump.toString());
|
---|
118 | // parses the fields and body text of a webpage in nutch's txt dump of entire site
|
---|
119 | //page.parseFields();
|
---|
120 | //page.getText();
|
---|
121 | pages.add(page);
|
---|
122 | pageDump = null;
|
---|
123 |
|
---|
124 | }
|
---|
125 |
|
---|
126 | // begin new webpage dump
|
---|
127 | pageDump = new StringBuilder();
|
---|
128 | pageDump.append(line);
|
---|
129 | pageDump.append("\n");
|
---|
130 |
|
---|
131 | }
|
---|
132 | else if(!line.equals("")) {
|
---|
133 | pageDump.append(line);
|
---|
134 | pageDump.append("\n");
|
---|
135 |
|
---|
136 | }
|
---|
137 | // can throw away any newlines between text start and end markers.
|
---|
138 |
|
---|
139 | prevLine = line;
|
---|
140 | }
|
---|
141 |
|
---|
142 | // process final webpage record:
|
---|
143 | //debugPageDump(pageDump);
|
---|
144 |
|
---|
145 | TextDumpPage page = new TextDumpPage(siteID, pageDump.toString());
|
---|
146 | pages.add(page);
|
---|
147 | pageDump = null;
|
---|
148 |
|
---|
149 | } catch (IOException ioe) {
|
---|
150 | error("@@@@@@@@@ Error reading in nutch txtdump file " + txtDumpFile, ioe);
|
---|
151 | }
|
---|
152 |
|
---|
153 | // Just do this once: get and store domain of site.
|
---|
154 | // Passing true to get domain with protocol prefix
|
---|
155 | if(pages.size() > 0) {
|
---|
156 | TextDumpPage firstPage = pages.get(0);
|
---|
157 | String url = firstPage.getPageURL();
|
---|
158 | this.domainOfSite = CCWETProcessor.getDomainForURL(url, true);
|
---|
159 | }
|
---|
160 | else {
|
---|
161 | this.domainOfSite = "UNKNOWN";
|
---|
162 | }
|
---|
163 |
|
---|
164 | prepareSiteStats();
|
---|
165 | }
|
---|
166 |
|
---|
167 | /** pageID: id into pages array */
|
---|
168 | public boolean isPageInMaori(int pageID) throws ArrayIndexOutOfBoundsException {
|
---|
169 |
|
---|
170 | String text = getTextForPage(pageID);
|
---|
171 |
|
---|
172 | // QTODO: what to do when page body text is empty?
|
---|
173 | if(text.equals("")) return false;
|
---|
174 | return maoriTxtDetector.isTextInMaori(text);
|
---|
175 | }
|
---|
176 |
|
---|
177 | private TextDumpPage getPage(int pageID) throws ArrayIndexOutOfBoundsException {
|
---|
178 | if(pageID < 0 || pageID >= pages.size()) {
|
---|
179 | throw new ArrayIndexOutOfBoundsException();
|
---|
180 | }
|
---|
181 |
|
---|
182 | TextDumpPage page = pages.get(pageID);
|
---|
183 | return page;
|
---|
184 | }
|
---|
185 |
|
---|
186 | public String getTextForPage(int pageID) throws ArrayIndexOutOfBoundsException {
|
---|
187 | TextDumpPage page = getPage(pageID);
|
---|
188 | return page.getPageText();
|
---|
189 | }
|
---|
190 | public String getURLForPage(int pageID) throws ArrayIndexOutOfBoundsException {
|
---|
191 | TextDumpPage page = getPage(pageID);
|
---|
192 | return page.getPageURL();
|
---|
193 | }
|
---|
194 |
|
---|
195 | public int totalNumPages() {
|
---|
196 | return pages.size();
|
---|
197 | }
|
---|
198 |
|
---|
199 | private void prepareSiteStats() {
|
---|
200 | pagesInMRI = new ArrayList<MRIWebPageStats>();
|
---|
201 | pagesContainingMRI = new ArrayList<MRIWebPageStats>();
|
---|
202 |
|
---|
203 | TextDumpPage page = null;
|
---|
204 | for(int i = 0; i < pages.size(); i++) {
|
---|
205 |
|
---|
206 | page = pages.get(i);
|
---|
207 |
|
---|
208 | String text = page.getPageText();
|
---|
209 | if(text.equals("")) {
|
---|
210 | page.addMRILanguageStatus(false);
|
---|
211 | continue;
|
---|
212 | }
|
---|
213 | else {
|
---|
214 | boolean isMRI = maoriTxtDetector.isTextInMaori(text);
|
---|
215 |
|
---|
216 | page.addMRILanguageStatus(isMRI);
|
---|
217 |
|
---|
218 | // Even if the entire page is not found to be overall in MÄori,
|
---|
219 | // let's sitll inspect the sentences of the page and count how many (if any)
|
---|
220 | // are in te reo.
|
---|
221 | ArrayList<String> mriSentences = maoriTxtDetector.getAllSentencesInMaori(text);
|
---|
222 | // first element is always total num sentences
|
---|
223 | // remaining elements are the actual sentences that were detected as being MÄori
|
---|
224 | int totalSentences = Integer.parseInt(mriSentences.get(0));
|
---|
225 | int numSentencesInMRI = mriSentences.size() - 1;
|
---|
226 |
|
---|
227 |
|
---|
228 | // Add page to list of MRI pages if the page's body text overall was detected
|
---|
229 | // as MÄori
|
---|
230 | // Add page to list of pages containing MRI if >= 1 sentences in the page
|
---|
231 | // were detected as being in MRI
|
---|
232 | if(isMRI || numSentencesInMRI >= 1) {
|
---|
233 | String url = page.getPageURL();
|
---|
234 | MRIWebPageStats MRIpageStats = new MRIWebPageStats(this.siteID, url, i, isMRI,
|
---|
235 | totalSentences, numSentencesInMRI);
|
---|
236 | if(isMRI) {
|
---|
237 | pagesInMRI.add(MRIpageStats);
|
---|
238 | } else if(numSentencesInMRI >= 1) {
|
---|
239 | pagesContainingMRI.add(MRIpageStats);
|
---|
240 | }
|
---|
241 |
|
---|
242 | }
|
---|
243 | }
|
---|
244 | }
|
---|
245 | }
|
---|
246 |
|
---|
247 | public void printSiteStats() {
|
---|
248 |
|
---|
249 |
|
---|
250 | info("------------- " + this.siteID + " SITE STATS -----------");
|
---|
251 |
|
---|
252 | info("SITE DOMAIN: " + this.domainOfSite);
|
---|
253 | info("Total number of web pages in site: " + pages.size());
|
---|
254 | info("Of these, the number of pages in MÄori (mri) were: " + this.pagesInMRI.size());
|
---|
255 |
|
---|
256 | if(pagesInMRI.size() > 0) {
|
---|
257 | info("The following were the pages detected by OpenNLP as being in MÄori with " + maoriTxtDetector.MINIMUM_CONFIDENCE + " confidence");
|
---|
258 | for(MRIWebPageStats mriWebPageInfo : pagesInMRI) {
|
---|
259 | info(mriWebPageInfo.toString());
|
---|
260 | }
|
---|
261 | }
|
---|
262 |
|
---|
263 | info(" ----------- ");
|
---|
264 | if(pagesContainingMRI.size() > 0) {
|
---|
265 | info("The following pages weren't detected as primarily being in MÄori");
|
---|
266 | info("But still contained sentences detected as MÄori");
|
---|
267 | for(MRIWebPageStats mriWebPageInfo : pagesContainingMRI) {
|
---|
268 | info(mriWebPageInfo.toString());
|
---|
269 | }
|
---|
270 |
|
---|
271 | } else {
|
---|
272 | info("No further pages detected as containing any sentences in MRI");
|
---|
273 | }
|
---|
274 | info(" ----------- ");
|
---|
275 | }
|
---|
276 |
|
---|
277 |
|
---|
278 | // --------------- STATIC METHODS AND INNER CLASSED USED BY MAIN -------------- //
|
---|
279 | public static void info(String msg) {
|
---|
280 | System.err.println(msg);
|
---|
281 | logger.info(msg);
|
---|
282 | }
|
---|
283 | public static void debug(String msg) {
|
---|
284 | System.err.println(msg);
|
---|
285 | logger.debug(msg);
|
---|
286 | }
|
---|
287 | public static void warn(String msg) {
|
---|
288 | System.err.println(msg);
|
---|
289 | logger.warn(msg);
|
---|
290 | }
|
---|
291 | public static void error(String msg) {
|
---|
292 | System.err.println(msg);
|
---|
293 | logger.error(msg);
|
---|
294 | }
|
---|
295 | public static void error(String msg, Exception e) {
|
---|
296 | logger.error(msg, e);
|
---|
297 | System.err.println("\n"+msg);
|
---|
298 | e.printStackTrace();
|
---|
299 | }
|
---|
300 |
|
---|
301 | public static void printUsage() {
|
---|
302 | info("Run this program as:");
|
---|
303 | info("\tNutchTextDumpProcessor <path to 'sites' folder>");
|
---|
304 | }
|
---|
305 |
|
---|
306 | public static void main(String[] args) {
|
---|
307 | if(args.length != 1) {
|
---|
308 | printUsage();
|
---|
309 | return;
|
---|
310 | }
|
---|
311 |
|
---|
312 | File sitesDir = new File(args[0]);
|
---|
313 | if(!sitesDir.exists() || !sitesDir.isDirectory()) {
|
---|
314 | error("Error: " + args[0] + " does not exist or is not a directory");
|
---|
315 | return;
|
---|
316 | }
|
---|
317 |
|
---|
318 | NutchTextDumpProcessor.DEBUG_MODE = false;
|
---|
319 |
|
---|
320 | try {
|
---|
321 | MaoriTextDetector mriTxtDetector = new MaoriTextDetector(true); // true: run silent
|
---|
322 | File[] sites = sitesDir.listFiles();
|
---|
323 |
|
---|
324 | // sort site folders in alphabetical order
|
---|
325 | // https://stackoverflow.com/questions/7199911/how-to-file-listfiles-in-alphabetical-order
|
---|
326 | Arrays.sort(sites);
|
---|
327 |
|
---|
328 | for(File siteDir : sites) { // e.g. 00001
|
---|
329 | if(siteDir.isDirectory()) {
|
---|
330 | // look for dump.txt
|
---|
331 | File txtDumpFile = new File(siteDir, "dump.txt");
|
---|
332 | if(!txtDumpFile.exists()) {
|
---|
333 | error("Text dump file " + txtDumpFile + " did not exist");
|
---|
334 | continue;
|
---|
335 | }
|
---|
336 |
|
---|
337 | else {
|
---|
338 | String siteID = siteDir.getName();
|
---|
339 | debug("Found siteID: " + siteID);
|
---|
340 | NutchTextDumpProcessor nutchTxtDump = new NutchTextDumpProcessor(mriTxtDetector, siteID, txtDumpFile);
|
---|
341 | // now it's parsed all the web pages in the site's text dump
|
---|
342 |
|
---|
343 | // Let's print stats on each web page's detected language being MRI or not
|
---|
344 | // and how many pages there were in the site in total.
|
---|
345 |
|
---|
346 | nutchTxtDump.printSiteStats();
|
---|
347 | }
|
---|
348 | }
|
---|
349 |
|
---|
350 | }
|
---|
351 |
|
---|
352 | } catch(Exception e) {
|
---|
353 | // can get an exception when instantiating CCWETProcessor instance
|
---|
354 | error(e.getMessage(), e);
|
---|
355 | }
|
---|
356 | }
|
---|
357 | }
|
---|