1 | package org.greenstone.atea;
|
---|
2 |
|
---|
3 | import java.util.*;
|
---|
4 | import java.io.*;
|
---|
5 |
|
---|
6 | import org.apache.log4j.Logger;
|
---|
7 |
|
---|
8 | /**
|
---|
9 | * Runs some of the important mongoDB queries I ran.
|
---|
10 | *
|
---|
11 | * TO COMPILE OR RUN, FIRST DO:
|
---|
12 | * cd maori-lang-detection/apache-opennlp-1.9.1
|
---|
13 | * export OPENNLP_HOME=`pwd`
|
---|
14 | * cd maori-lang-detection/src
|
---|
15 | *
|
---|
16 | * TO COMPILE:
|
---|
17 | * maori-lang-detection/src$
|
---|
18 | * javac -cp ".:../conf:../lib/*" org/greenstone/atea/SummaryTool.java
|
---|
19 | *
|
---|
20 | * TO RUN:
|
---|
21 | * maori-lang-detection/src$
|
---|
22 | * java -cp ".:../conf:../lib/*" org/greenstone/atea/SummaryTool
|
---|
23 | * OR:
|
---|
24 | * java -cp ".:../conf:../lib/*" org/greenstone/atea/SummaryTool ../mongodb-data/domainsNZ_IsMRI.txt
|
---|
25 | *
|
---|
26 | */
|
---|
27 | public class SummaryTool {
|
---|
28 | static Logger logger = Logger.getLogger(org.greenstone.atea.SummaryTool.class.getName());
|
---|
29 | static private final long FIXED_SEED = 1000;
|
---|
30 |
|
---|
31 | private final MongoDBQueryer mongodbQueryer;
|
---|
32 | private File outFolder;
|
---|
33 |
|
---|
34 |
|
---|
35 |
|
---|
36 | public static class Tuple {
|
---|
37 | public final String url;
|
---|
38 | public final String countryCode;
|
---|
39 |
|
---|
40 | public Tuple(String url, String countryCode) {
|
---|
41 | this.url = url;
|
---|
42 | this.countryCode = countryCode;
|
---|
43 | }
|
---|
44 |
|
---|
45 | public String toString() {
|
---|
46 | return this.url + "," + countryCode;
|
---|
47 | }
|
---|
48 | }
|
---|
49 |
|
---|
50 |
|
---|
51 | public SummaryTool(MongoDBQueryer mongodbQueryer, File outFolder)
|
---|
52 | {
|
---|
53 | this.mongodbQueryer = mongodbQueryer;
|
---|
54 | this.outFolder = outFolder;
|
---|
55 | }
|
---|
56 |
|
---|
57 |
|
---|
58 | public void produceURLsForPagesInMRI(File domainsFile) {
|
---|
59 | ArrayList<Tuple> urlsList = getURLsForAllWebPagesInSiteListing(MongoDBQueryer.IS_MRI, domainsFile);
|
---|
60 | File outFile = new File(outFolder, "isMRI_"+domainsFile.getName());
|
---|
61 | writeURLsToFile(urlsList, outFile, urlsList.size());
|
---|
62 |
|
---|
63 | System.out.println("Wrote all isMRI web page URLs for the sites in input domainsFile\ninto file: "
|
---|
64 | + Utility.getFilePath(outFile));
|
---|
65 | }
|
---|
66 |
|
---|
67 | public void produceURLsForPagesContainingMRI(File domainsFile) {
|
---|
68 | ArrayList<Tuple> urlsList = getURLsForAllWebPagesInSiteListing(MongoDBQueryer.CONTAINS_MRI, domainsFile);
|
---|
69 | File outFile = new File(outFolder, "containsMRI_"+domainsFile.getName());
|
---|
70 | writeURLsToFile(urlsList, outFile, urlsList.size());
|
---|
71 |
|
---|
72 | System.out.println("Wrote all containsMRI web page URLs for the sites in input domainsFile\ninto file: "
|
---|
73 | + Utility.getFilePath(outFile));
|
---|
74 | }
|
---|
75 |
|
---|
76 | private ArrayList<Tuple> getURLsForAllWebPagesInSiteListing(int filterType, File domainsFile) {
|
---|
77 | ArrayList<Tuple> urlsList = new ArrayList<Tuple>();
|
---|
78 |
|
---|
79 | // 1. read each url from the domainsFile
|
---|
80 | // 1a. do the query
|
---|
81 | // 1b. add the arraylist result to urls
|
---|
82 |
|
---|
83 | try (
|
---|
84 | BufferedReader reader = new BufferedReader(new FileReader(domainsFile));
|
---|
85 | ) {
|
---|
86 |
|
---|
87 | String domain;
|
---|
88 |
|
---|
89 | while((domain = reader.readLine()) != null) {
|
---|
90 | domain = domain.trim();
|
---|
91 | if(!domain.equals("")) {
|
---|
92 |
|
---|
93 | String countryCode = "";
|
---|
94 | int index = domain.lastIndexOf(",");
|
---|
95 | if(index != -1) {
|
---|
96 | countryCode = domain.substring(index+1).trim();
|
---|
97 | domain = domain.substring(0, index);
|
---|
98 | }
|
---|
99 | ArrayList<String> moreURLs = mongodbQueryer.queryAllMatchingURLsFilteredBy(domain, filterType);
|
---|
100 |
|
---|
101 | // Print out whether there were no isMRI pages for the domain (only containsMRI). A useful thing to know
|
---|
102 | if(moreURLs.size() == 0 && filterType == MongoDBQueryer.IS_MRI) {
|
---|
103 | System.out.println(" " + countryCode + " domain " + domain + " had no isMRI webpages - only containsMRI.");
|
---|
104 | }
|
---|
105 |
|
---|
106 | //urlsList.addAll(moreURLs);
|
---|
107 | for(int i = 0; i < moreURLs.size(); i++) {
|
---|
108 | urlsList.add(new Tuple(moreURLs.get(i), countryCode));
|
---|
109 | }
|
---|
110 |
|
---|
111 | }
|
---|
112 | }
|
---|
113 | System.err.println("");
|
---|
114 | } catch(Exception e) {
|
---|
115 | logger.error("Unable to read URLs from file " + Utility.getFilePath(domainsFile));
|
---|
116 | logger.error(e.getMessage(), e);
|
---|
117 | }
|
---|
118 |
|
---|
119 | return urlsList;
|
---|
120 | }
|
---|
121 |
|
---|
122 | /** Given a hand curated list of all sites with positive numPagesContainingMRI
|
---|
123 | * determined by manual inspection, get a listing of all their web pages that
|
---|
124 | * are IN_MRI (or CONTAINS_MRI?).
|
---|
125 | * Total all these pages that are inMRI (N), then work out the correct sample size (n)
|
---|
126 | * at 90% confidence with 5% margin of error. Then generate a random listing
|
---|
127 | * of n of these pages in MRI of these trusted sites and output to a file
|
---|
128 | * for manual inspection of the sample webpage URLs at page-level. */
|
---|
129 | /* OLD: Given a hand curated list of non-NZ sites that CONTAINS_MRI, get a listing
|
---|
130 | * of all their web pages IN_MRI (or CONTAINS_MRI).
|
---|
131 | * Plus a listing of all the NZ pages IN_MRI. */
|
---|
132 | //public void webPagesOfAllNZSitesAndDomainListing(File domainsFile) {
|
---|
133 | public void mriWebPageListingForDomainListing(File domainsFile) {
|
---|
134 |
|
---|
135 | int filterType = MongoDBQueryer.IS_MRI;
|
---|
136 |
|
---|
137 | // for overseas websites,
|
---|
138 | //produceURLsForPagesContainingMRI(handCuratedOverseasDomainsFile);
|
---|
139 |
|
---|
140 | // 0. get a list of all the web pages in the given domain listing where isMRI = true
|
---|
141 | ArrayList<Tuple> urlsList = getURLsForAllWebPagesInSiteListing(MongoDBQueryer.IS_MRI, domainsFile);
|
---|
142 | // produceURLsForPagesInMRI(domainsFile);
|
---|
143 |
|
---|
144 | // 1. calculate the population size, N, the number of all webpages in the given domain
|
---|
145 | // site listing where isMRI = true.
|
---|
146 | int N_totalNumPages = urlsList.size();
|
---|
147 |
|
---|
148 | // 2. write all the URLs in urlsList to a file
|
---|
149 | //File outFolder = domainsFile.getParentFile();
|
---|
150 | String fileName = (filterType == MongoDBQueryer.IS_MRI) ? "isMRI_" : "containsMRI_";
|
---|
151 | File outFile = new File(outFolder, fileName+domainsFile.getName());
|
---|
152 |
|
---|
153 | writeURLsToFile(urlsList, outFile, N_totalNumPages);
|
---|
154 | System.out.println("Wrote out full listing of web page URLs for sites in input domainsFile"
|
---|
155 | + "\ninto file: " + Utility.getFilePath(outFile));
|
---|
156 |
|
---|
157 | // 3. calculate sample size n for population size N if using 90% confidence and 5% margin of error
|
---|
158 | int n_numSampleURLs = calcSampleSize(N_totalNumPages);
|
---|
159 |
|
---|
160 | System.err.println("*** N, total number of web pages that matched: " + N_totalNumPages);
|
---|
161 | System.err.println("*** n, sample size of web page URLs: " + n_numSampleURLs);
|
---|
162 |
|
---|
163 | // 4. Shuffle all the URLs and write the first n (sample size) URLs to a file
|
---|
164 | // Using a constant seed for reproducibility
|
---|
165 | // https://stackoverflow.com/questions/6284589/setting-a-seed-to-shuffle-arraylist-in-java-deterministically
|
---|
166 | Collections.shuffle(urlsList, new Random(FIXED_SEED));
|
---|
167 |
|
---|
168 | outFile = new File(outFolder, "random"+n_numSampleURLs+"_"+domainsFile.getName());
|
---|
169 | writeURLsToFile(urlsList, outFile, n_numSampleURLs);
|
---|
170 | System.out.println("Wrote a sample of n=" + n_numSampleURLs + " of web page URLs "
|
---|
171 | + "for the sites in input domainsFile\ninto file: " + Utility.getFilePath(outFile));
|
---|
172 | }
|
---|
173 |
|
---|
174 | /**
|
---|
175 | * Calculates sample size n for binary outcomes at 90% confidence and 5% margine of error
|
---|
176 | * for given population size N.
|
---|
177 | * @return n, the sample size.
|
---|
178 | */
|
---|
179 | public int calcSampleSize(int N) {
|
---|
180 |
|
---|
181 | // calculate sample size n for population size N if using 90% confidence and 5% margin of error
|
---|
182 | // https://stats.stackexchange.com/questions/207584/sample-size-choice-with-binary-outcome
|
---|
183 | // https://www.statisticshowto.datasciencecentral.com/probability-and-statistics/find-sample-size/#CI1
|
---|
184 | // https://www.statisticshowto.datasciencecentral.com/z-alpha2-za2/
|
---|
185 |
|
---|
186 | double m = 0.05; // margin of error = 5%
|
---|
187 | // for 90% confidence, alpha is the remainder = 10% and alpha/2 = 5%.
|
---|
188 | // For 90% confidence, use the table of known z_alpha/2 values from step 1 of
|
---|
189 | // https://www.statisticshowto.datasciencecentral.com/z-alpha2-za2/
|
---|
190 | double z_alpha_over_2 = 1.6449;
|
---|
191 |
|
---|
192 | // Formula: n = (zalpha2 ^ 2 * N) / ((z-alpha-2 ^ 2) + 4(N-1)*m^2)
|
---|
193 | // see https://stats.stackexchange.com/questions/207584/sample-size-choice-with-binary-outcome
|
---|
194 | double n = (Math.pow(z_alpha_over_2, 2.0) * N) / (Math.pow(z_alpha_over_2, 2.0) + (4 * (N-1) * Math.pow(m,2.0)));
|
---|
195 |
|
---|
196 | // Round up to get a whole number:
|
---|
197 | return (int)Math.ceil(n);
|
---|
198 | }
|
---|
199 |
|
---|
200 | /**
|
---|
201 | * Writes out the first n URLs in urlsList into outFile.
|
---|
202 | */
|
---|
203 | private void writeURLsToFile(ArrayList<Tuple> urlsList, File outFile, final int n) {
|
---|
204 | try (
|
---|
205 | Writer writer = new BufferedWriter(new FileWriter(outFile));
|
---|
206 | ) {
|
---|
207 |
|
---|
208 | for (int i=0; i < n; i++) {
|
---|
209 | Tuple urlInfo = urlsList.get(i);
|
---|
210 |
|
---|
211 | //System.out.println(list.get(i));
|
---|
212 | writer.write(urlInfo + "\n"); // calls toString() on tuple of url -> countryCode
|
---|
213 | }
|
---|
214 | } catch(Exception e) {
|
---|
215 | logger.error("Unable to write to file " + Utility.getFilePath(outFile));
|
---|
216 | logger.error(e.getMessage(), e);
|
---|
217 | }
|
---|
218 | }
|
---|
219 |
|
---|
220 | /* ---------------------------------------- */
|
---|
221 | /**
|
---|
222 | * Create the file 5counts_containsMRISites_allNZGrouped.json
|
---|
223 | * that contains the count and domains for NZ sites (NZ origin or nz TLD) with pages
|
---|
224 | * that CONTAIN_MRI, followed by counts and domains listing for overseas sites
|
---|
225 | * that CONTAIN_MRI.
|
---|
226 | * @return full path of file generated
|
---|
227 | */
|
---|
228 | public String writeContainsMRISites_nzSitesAndTLDsGrouped() {
|
---|
229 |
|
---|
230 | File outFile = new File(outFolder, "5counts_containsMRISites_allNZGrouped.json");
|
---|
231 |
|
---|
232 | String filename = Utility.getFilePath(outFile);
|
---|
233 |
|
---|
234 | try (
|
---|
235 | Writer writer = new BufferedWriter(new FileWriter(outFile));
|
---|
236 | ) {
|
---|
237 | // first write out NZ sites and .nz TLD count and domains
|
---|
238 | mongodbQueryer.aggregateContainsMRIForNZ(writer, MongoDBQueryer.CONTAINS_MRI);
|
---|
239 | // next write out all overseas sites (not NZ origin or .nz TLD)
|
---|
240 | // that have no "mi" in the URL path as mi.* or */mi
|
---|
241 | boolean isMiInURLPath = false;
|
---|
242 | mongodbQueryer.aggregateContainsMRIForOverseas(writer, MongoDBQueryer.CONTAINS_MRI);
|
---|
243 |
|
---|
244 | } catch(Exception e) {
|
---|
245 | logger.error("Unable to write to file " + filename);
|
---|
246 | logger.error(e.getMessage(), e);
|
---|
247 | }
|
---|
248 |
|
---|
249 | System.err.println("*** Wrote file: " + filename);
|
---|
250 |
|
---|
251 | return filename;
|
---|
252 | }
|
---|
253 |
|
---|
254 | /**
|
---|
255 | * Create the file 5a_counts_tentativeNonAutotranslatedSites.json
|
---|
256 | * that contains the count and domains for NZ sites (NZ origin or nz TLD) that CONTAIN_MRI
|
---|
257 | * followed by counts and domain listing for overseas sites that are either from Australia
|
---|
258 | * or don't contain mi in their URL path.
|
---|
259 | * @return full path of file generated
|
---|
260 | */
|
---|
261 | public String writeTentativeNonAutotranslatedSites() {
|
---|
262 |
|
---|
263 | File outFile = new File(outFolder, "5a_counts_tentativeNonAutotranslatedSites.json");
|
---|
264 |
|
---|
265 | String filename = Utility.getFilePath(outFile);
|
---|
266 |
|
---|
267 | try (
|
---|
268 | Writer writer = new BufferedWriter(new FileWriter(outFile));
|
---|
269 | ) {
|
---|
270 | // first write out NZ sites and .nz TLD count and domains
|
---|
271 | mongodbQueryer.aggregateContainsMRIForNZ(writer, MongoDBQueryer.CONTAINS_MRI);
|
---|
272 | // next write out all overseas sites (not NZ origin or .nz TLD)
|
---|
273 | // that have no "mi" in the URL path as mi.* or */mi
|
---|
274 | boolean isMiInURLPath = false;
|
---|
275 | mongodbQueryer.aggregateContainsMRIForOverseas(writer, MongoDBQueryer.CONTAINS_MRI, isMiInURLPath);
|
---|
276 |
|
---|
277 | } catch(Exception e) {
|
---|
278 | logger.error("Unable to write to file " + filename);
|
---|
279 | logger.error(e.getMessage(), e);
|
---|
280 | }
|
---|
281 |
|
---|
282 | System.err.println("*** Wrote file: " + filename);
|
---|
283 |
|
---|
284 | return filename;
|
---|
285 | }
|
---|
286 |
|
---|
287 | /**
|
---|
288 | * Create the file 5b_counts_overseasSitesWithMiInPath.json
|
---|
289 | * Listing of the remainder of overseas sites that CONTAIN_MRI not included by
|
---|
290 | * writeTentativeNonAutotranslatedSites(): those that have mi in their URL path.
|
---|
291 | * This listing is separate to allow easier weeding out of product sites/autotranslated
|
---|
292 | * sites when eyeballing the listing output.
|
---|
293 | */
|
---|
294 | public String writeOverseasSitesWithMiInURLPath() {
|
---|
295 | File outFile = new File(outFolder, "5b_counts_overseasSitesWithMiInPath.json");
|
---|
296 |
|
---|
297 | String filename = Utility.getFilePath(outFile);
|
---|
298 | try (
|
---|
299 | Writer writer = new BufferedWriter(new FileWriter(outFile));
|
---|
300 | ) {
|
---|
301 | boolean isMiInURLPath = true;
|
---|
302 | mongodbQueryer.aggregateContainsMRIForOverseas(writer, MongoDBQueryer.CONTAINS_MRI, isMiInURLPath);
|
---|
303 |
|
---|
304 | } catch(Exception e) {
|
---|
305 | logger.error("Unable to write to file " + filename);
|
---|
306 | logger.error(e.getMessage(), e);
|
---|
307 | }
|
---|
308 |
|
---|
309 | System.err.println("*** Wrote file: " + filename);
|
---|
310 | return filename;
|
---|
311 | }
|
---|
312 |
|
---|
313 | public static void printUsage() {
|
---|
314 | System.err.println("Usage: SummaryTool [domains.txt]");
|
---|
315 | }
|
---|
316 |
|
---|
317 | /**
|
---|
318 | * If no args are passed in, generates complete containsMRI file listings for NZ and overseas web SITES (domains),
|
---|
319 | * with overseas web sites that have mi (mi.* or *\/mi) in the URL path listed separately.
|
---|
320 | * You can then manually inspect the domains in this listing to shortlist which of these sites are not automatically
|
---|
321 | * translated and really contain at least one webpage containing at least one sentence in MRI.
|
---|
322 | * If a file is passed in containing a list of domains, then this first generates a full listing of all webpages
|
---|
323 | * matching isMRI for each site in the domain list. It then generates a smaller set of random webpages matching
|
---|
324 | * isMRI for the pooled sites in the domain list where the sample size of URLs produced is sufficient for giving
|
---|
325 | * 90% confidence with 5% margin of error for testing binary outcomes, see
|
---|
326 | * https://stats.stackexchange.com/questions/207584/sample-size-choice-with-binary-outcome
|
---|
327 | */
|
---|
328 | public static void main(String args[]) {
|
---|
329 | if(args.length >= 2) {
|
---|
330 | printUsage();
|
---|
331 | System.exit(-1);
|
---|
332 | }
|
---|
333 |
|
---|
334 | try (
|
---|
335 | MongoDBQueryer mongodb = new MongoDBQueryer();
|
---|
336 | ) {
|
---|
337 |
|
---|
338 | mongodb.connectToDB();
|
---|
339 |
|
---|
340 | // output files will be stored in mongodb-data-auto
|
---|
341 | File outFolder = new File("../mongodb-data-auto/").getAbsoluteFile();
|
---|
342 | SummaryTool listing = new SummaryTool(mongodb, outFolder);
|
---|
343 |
|
---|
344 | System.out.println("*************************************");
|
---|
345 |
|
---|
346 |
|
---|
347 | if(args.length >= 1) { // web page urls listing mode
|
---|
348 | File domainsFile = new File(args[0]);
|
---|
349 | if(!domainsFile.exists()) {
|
---|
350 | System.err.println("File " + domainsFile + " does not exist");
|
---|
351 | System.exit(-1);
|
---|
352 | }
|
---|
353 |
|
---|
354 | //String isMRIFile = listing.produceURLsForPagesInMRI(domainsFile);
|
---|
355 | //String containsMRIFile = listing.produceURLsForPagesContainingMRI(domainsFile);
|
---|
356 |
|
---|
357 |
|
---|
358 | // TODO: for NZ, do IS_MRI. For overseas still CONTAINS_MRI
|
---|
359 | // then also do the shuffle to gen X num of random web page URLs.
|
---|
360 | //String filename = listing.webPagesOfAllNZSitesAndDomainListing(domainsFile);
|
---|
361 | listing.mriWebPageListingForDomainListing(domainsFile);
|
---|
362 |
|
---|
363 | // TODO: generate the special table (6)
|
---|
364 |
|
---|
365 | } else {
|
---|
366 |
|
---|
367 | // calculating sample size works:
|
---|
368 | //System.err.println("For N = " + 4360 + ", n = " + listing.calcSampleSize(4360));
|
---|
369 | //System.err.println("For N = " + 681 + ", n = " + listing.calcSampleSize(681));
|
---|
370 |
|
---|
371 | // get all sites where >0 pages have containsMRI=true
|
---|
372 | // grouping NZ sites and .nz TLDs together and remainder under overseas
|
---|
373 | // geolocations.
|
---|
374 | String filename = listing.writeContainsMRISites_nzSitesAndTLDsGrouped();
|
---|
375 |
|
---|
376 | // separately:
|
---|
377 | // - all NZ containsMRI + overseas tentative non-product sites with containMRI
|
---|
378 | // - overseas tentative product sites with containMRI
|
---|
379 | filename = listing.writeTentativeNonAutotranslatedSites();
|
---|
380 | filename = listing.writeOverseasSitesWithMiInURLPath();
|
---|
381 |
|
---|
382 | // TODO: generate the tables
|
---|
383 |
|
---|
384 | mongodb.writeTables(outFolder);
|
---|
385 | }
|
---|
386 |
|
---|
387 | System.out.println("*************************************");
|
---|
388 | } catch(Exception e) {
|
---|
389 | logger.error(e.getMessage(), e);
|
---|
390 | }
|
---|
391 | }
|
---|
392 | }
|
---|