File size: 22,816 Bytes
18573e4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 | package bg.bas.dcl.LLMs.IfGPTDataset;
import java.io.File;
import java.io.FileOutputStream;
import java.io.OutputStreamWriter;
import java.io.PrintWriter;
import java.io.Writer;
import java.nio.file.Files;
import java.nio.file.StandardCopyOption;
import java.util.ArrayList;
import java.util.Collections;
import java.util.HashMap;
import java.util.HashSet;
import java.util.LinkedHashMap;
import java.util.List;
import java.util.Map;
import java.util.Scanner;
import java.util.Set;
import java.util.TreeSet;
import info.debatty.java.lsh.MinHash;
import bg.bas.dcl.general.FileHandler;
/**
* DeduplicationProcessor — sentence-level near-duplicate detection
* using MinHash + LSH (Jaccard similarity).
*
* -----------------------------------------------------------------------
* MAVEN DEPENDENCY (add to pom.xml):
*
* <dependency>
* <groupId>info.debatty</groupId>
* <artifactId>java-lsh</artifactId>
* <version>0.12</version>
* </dependency>
*
* -----------------------------------------------------------------------
* HOW IT WORKS
*
* 1. INDEX phase — reads all .txt files in the "full corpus" directory.
* Each sentence is shingled into character n-grams, converted to a
* boolean vector over a shared vocabulary, and a MinHash signature
* is computed. All signatures are stored in an in-memory index keyed
* by (file, lineNumber).
*
* 2. QUERY phase — reads every sentence in the "new folder".
* For each sentence its MinHash signature is compared against every
* indexed corpus signature (approximate Jaccard via signature similarity).
* Pairs whose estimated Jaccard similarity ≥ threshold are reported.
*
* 3. REPORT — a TSV report is written listing every duplicate pair:
* new-file | new-line | corpus-file | corpus-line | similarity | sentence
*
* 4. OPTIONAL REMOVE — sentences in the new folder that are duplicates of
* corpus sentences are stripped from their file (originals backed up).
* Files that become empty after removal are deleted.
*
* -----------------------------------------------------------------------
* PARAMETERS
*
* threshold — Jaccard similarity to call a near-duplicate (default 0.90)
* shingleSize — character n-gram size for shingling (default 5)
* numHashes — number of hash functions for MinHash (default 200)
* More hashes → better accuracy, slower index.
*
* -----------------------------------------------------------------------
* USAGE
*
* DeduplicationProcessor dp = new DeduplicationProcessor(0.90);
* dp.indexCorpus("/path/to/full/corpus/");
* dp.detectDuplicates("/path/to/new/folder/", "/path/to/report.tsv");
* dp.removeDuplicatesFromNewFolder("/path/to/new/folder/", true); // true=keep .bak
*/
public class DeduplicationProcessor {
// -----------------------------------------------------------------------
// Configuration
// -----------------------------------------------------------------------
private final double threshold; // Jaccard similarity cut-off
private final int shingleSize; // character n-gram size
private final int numHashes; // MinHash signature length
// -----------------------------------------------------------------------
// Index state (built during indexCorpus)
// -----------------------------------------------------------------------
/** Shared vocabulary: every distinct shingle seen across all corpus sentences. */
private final Set<String> vocabulary = new HashSet<>();
/**
* Corpus index: maps SentenceKey → raw sentence text + MinHash signature.
* Built in two passes to allow vocabulary to be finalised before signing.
*/
private final Map<SentenceKey, IndexedSentence> corpusIndex = new LinkedHashMap<>();
/** MinHash object — initialised once vocabulary size is known. */
private MinHash minHash;
// -----------------------------------------------------------------------
// Duplicate results (populated by detectDuplicates)
// -----------------------------------------------------------------------
/** All duplicate pairs found in the last detectDuplicates run. */
private final List<DuplicatePair> duplicatePairs = new ArrayList<>();
/**
* Set of SentenceKeys in the NEW folder that are duplicates.
* Used by removeDuplicatesFromNewFolder.
*/
private final Set<SentenceKey> duplicateNewSentences = new HashSet<>();
// -----------------------------------------------------------------------
// Constructor
// -----------------------------------------------------------------------
public DeduplicationProcessor(double threshold) {
this(threshold, 5, 200);
}
public DeduplicationProcessor(double threshold, int shingleSize, int numHashes) {
if (threshold < 0 || threshold > 1)
throw new IllegalArgumentException("Threshold must be in [0, 1].");
this.threshold = threshold;
this.shingleSize = shingleSize;
this.numHashes = numHashes;
}
// -----------------------------------------------------------------------
// Phase 1 — Index the full corpus
// -----------------------------------------------------------------------
/**
* Reads all .txt files in {@code corpusDir}, shingles every sentence,
* builds a shared vocabulary, and computes MinHash signatures.
*
* This must be called before {@link #detectDuplicates}.
*
* @param corpusDir directory of clean .txt files representing the full corpus
*/
public void indexCorpus(String corpusDir) {
System.out.println("[Index] Scanning corpus: " + corpusDir);
try {
FileHandler fh = new FileHandler();
// --- Pass 1: collect sentences and build vocabulary ---
// Temporary store: key → raw text + shingle set (signatures computed later)
Map<SentenceKey, Set<String>> rawShingles = new LinkedHashMap<>();
for (File f : fh.getFileListing(new File(corpusDir))) {
if (!f.isFile() || !f.getName().endsWith(".txt")) continue;
Scanner sc = new Scanner(f, "UTF-8");
int lineNum = 0;
while (sc.hasNextLine()) {
String line = sc.nextLine().trim();
lineNum++;
if (line.length() < shingleSize) continue;
Set<String> shingles = shingle(line);
vocabulary.addAll(shingles);
rawShingles.put(new SentenceKey(f.getName(), lineNum), shingles);
}
sc.close();
}
System.out.println("[Index] Vocabulary size: " + vocabulary.size()
+ " Sentences: " + rawShingles.size());
if (vocabulary.isEmpty()) {
System.err.println("[Index] No sentences found — aborting.");
return;
}
// --- Initialise MinHash with finalised vocabulary size ---
// Error parameter 0.05 → ~400 hashes needed; we use numHashes directly.
// The debatty MinHash constructor accepts (error, dictSize).
// We use the lower-level approach: fix numHashes via the signature size.
// info.debatty MinHash(double error, int dictSize) chooses hash count itself.
// For explicit control we pass a small error so it aligns with numHashes.
minHash = new MinHash(numHashes, vocabulary.size());
// --- Pass 2: compute and store signatures ---
List<String> vocabList = new ArrayList<>(vocabulary);
corpusIndex.clear();
// Also keep a raw-text map for the report
Map<SentenceKey, String> rawTexts = new HashMap<>();
// re-scan to get raw text (we only stored shingles above)
for (File f : fh.getFileListing(new File(corpusDir))) {
if (!f.isFile() || !f.getName().endsWith(".txt")) continue;
Scanner sc = new Scanner(f, "UTF-8");
int lineNum = 0;
while (sc.hasNextLine()) {
String line = sc.nextLine().trim();
lineNum++;
if (line.length() < shingleSize) continue;
rawTexts.put(new SentenceKey(f.getName(), lineNum), line);
}
sc.close();
}
for (Map.Entry<SentenceKey, Set<String>> entry : rawShingles.entrySet()) {
SentenceKey key = entry.getKey();
boolean[] vector = toVector(entry.getValue(), vocabList);
int[] sig = minHash.signature(vector);
String rawText = rawTexts.getOrDefault(key, "");
corpusIndex.put(key, new IndexedSentence(rawText, sig));
}
System.out.println("[Index] Corpus index built: "
+ corpusIndex.size() + " sentences.");
} catch (Exception e) {
e.printStackTrace();
}
}
// -----------------------------------------------------------------------
// Phase 2 — Detect duplicates in new folder
// -----------------------------------------------------------------------
/**
* Compares every sentence in {@code newDir} against the corpus index.
* Pairs with estimated Jaccard ≥ threshold are recorded as duplicates
* and written to {@code reportPath}.
*
* Call {@link #indexCorpus} first.
*
* @param newDir directory of new .txt files to check
* @param reportPath destination TSV report file
*/
public void detectDuplicates(String newDir, String reportPath) {
if (corpusIndex.isEmpty()) {
System.err.println("[Detect] Corpus index is empty. Call indexCorpus() first.");
return;
}
System.out.println("[Detect] Comparing new folder against corpus index...");
duplicatePairs.clear();
duplicateNewSentences.clear();
List<String> vocabList = new ArrayList<>(vocabulary);
try {
FileHandler fh = new FileHandler();
for (File f : fh.getFileListing(new File(newDir))) {
if (!f.isFile() || !f.getName().endsWith(".txt")) continue;
System.out.println("[Detect] Checking: " + f.getName());
Scanner sc = new Scanner(f, "UTF-8");
int lineNum = 0;
while (sc.hasNextLine()) {
String line = sc.nextLine().trim();
lineNum++;
if (line.length() < shingleSize) continue;
Set<String> shingles = shingle(line);
// Only shingles already in vocabulary are meaningful
Set<String> filtered = new HashSet<>(shingles);
filtered.retainAll(vocabulary);
// If almost none of the shingles are in vocab → skip
// (the sentence is likely from a very different domain)
if (filtered.isEmpty()) continue;
boolean[] newVec = toVector(filtered, vocabList);
int[] newSig = minHash.signature(newVec);
SentenceKey newKey = new SentenceKey(f.getName(), lineNum);
// Compare against all corpus sentences
// For large corpora, replace this loop with an LSH band index
for (Map.Entry<SentenceKey, IndexedSentence> entry : corpusIndex.entrySet()) {
double sim = minHash.similarity(newSig, entry.getValue().signature);
if (sim >= threshold) {
DuplicatePair pair = new DuplicatePair(
newKey, line,
entry.getKey(), entry.getValue().text,
sim);
duplicatePairs.add(pair);
duplicateNewSentences.add(newKey);
// Don't break — report ALL corpus matches for transparency
}
}
}
sc.close();
}
System.out.println("[Detect] Duplicate sentence pairs found: "
+ duplicatePairs.size());
System.out.println("[Detect] Unique new sentences flagged: "
+ duplicateNewSentences.size());
writeReport(reportPath);
} catch (Exception e) {
e.printStackTrace();
}
}
// -----------------------------------------------------------------------
// Phase 3 — Optionally remove duplicates from new folder
// -----------------------------------------------------------------------
/**
* Removes from every file in {@code newDir} any sentence whose
* (file, lineNumber) is in the duplicate set detected by
* {@link #detectDuplicates}.
*
* Files that become empty after removal are deleted.
* Must be called after {@link #detectDuplicates}.
*
* @param newDir directory of new .txt files to clean
* @param keepBackup if true, originals are renamed to *.bak first
*/
public void removeDuplicatesFromNewFolder(String newDir, boolean keepBackup) {
if (duplicateNewSentences.isEmpty()) {
System.out.println("[Remove] No duplicates to remove.");
return;
}
System.out.println("[Remove] Removing "
+ duplicateNewSentences.size() + " duplicate sentences...");
try {
FileHandler fh = new FileHandler();
int filesModified = 0;
int totalRemoved = 0;
for (File f : fh.getFileListing(new File(newDir))) {
if (!f.isFile() || !f.getName().endsWith(".txt")) continue;
List<String> inputLines = new ArrayList<>();
Scanner sc = new Scanner(f, "UTF-8");
int lineNum = 0;
while (sc.hasNextLine()) {
inputLines.add(sc.nextLine());
lineNum++;
}
sc.close();
List<String> outputLines = new ArrayList<>();
int removed = 0;
for (int i = 0; i < inputLines.size(); i++) {
String trimmed = inputLines.get(i).trim();
// +1 because lineNum was 1-based during indexing
SentenceKey key = new SentenceKey(f.getName(), i + 1);
if (trimmed.length() >= shingleSize
&& duplicateNewSentences.contains(key)) {
removed++;
} else {
outputLines.add(inputLines.get(i));
}
}
if (removed > 0) {
if (keepBackup) {
Files.copy(f.toPath(),
new File(f.getAbsolutePath() + ".bak").toPath(),
StandardCopyOption.REPLACE_EXISTING);
}
// Check if file would become empty (only blank lines)
boolean allBlank = outputLines.stream()
.allMatch(String::isBlank);
if (allBlank) {
f.delete();
System.out.println("[Remove] Deleted (empty after dedup): "
+ f.getName());
} else {
Writer w = new OutputStreamWriter(
new FileOutputStream(f), "UTF-8");
for (String l : outputLines) {
w.write(l + "\n");
}
w.flush();
w.close();
System.out.println("[Remove] " + f.getName()
+ " — removed " + removed + " sentences.");
}
filesModified++;
totalRemoved += removed;
}
}
System.out.println("[Remove] Done. Files modified: " + filesModified
+ " Sentences removed: " + totalRemoved);
} catch (Exception e) {
e.printStackTrace();
}
}
// -----------------------------------------------------------------------
// Report writer
// -----------------------------------------------------------------------
private void writeReport(String reportPath) throws Exception {
try (PrintWriter pw = new PrintWriter(
new OutputStreamWriter(new FileOutputStream(reportPath), "UTF-8"))) {
// Header
pw.println("# DeduplicationProcessor report");
pw.println("# Threshold: " + threshold
+ " ShingleSize: " + shingleSize
+ " NumHashes: " + numHashes);
pw.println("# Duplicate pairs: " + duplicatePairs.size());
pw.println("# Unique new sentences flagged: " + duplicateNewSentences.size());
pw.println();
pw.println("NEW_FILE\tNEW_LINE\tCORPUS_FILE\tCORPUS_LINE\tSIMILARITY\tNEW_SENTENCE\tCORPUS_SENTENCE");
// Sort by similarity descending, then new file, then line
List<DuplicatePair> sorted = new ArrayList<>(duplicatePairs);
sorted.sort((a, b) -> {
int cmp = Double.compare(b.similarity, a.similarity);
if (cmp != 0) return cmp;
cmp = a.newKey.fileName.compareTo(b.newKey.fileName);
if (cmp != 0) return cmp;
return Integer.compare(a.newKey.lineNumber, b.newKey.lineNumber);
});
for (DuplicatePair p : sorted) {
pw.printf("%s\t%d\t%s\t%d\t%.4f\t%s\t%s%n",
p.newKey.fileName,
p.newKey.lineNumber,
p.corpusKey.fileName,
p.corpusKey.lineNumber,
p.similarity,
sanitiseTsv(p.newText),
sanitiseTsv(p.corpusText));
}
}
System.out.println("[Report] Written to: " + reportPath);
}
// -----------------------------------------------------------------------
// Shingling and vectorisation helpers
// -----------------------------------------------------------------------
/**
* Produces the set of character n-grams (shingles) for a sentence.
* Lowercased so matching is case-insensitive.
*/
private Set<String> shingle(String text) {
Set<String> shingles = new TreeSet<>();
String lower = text.toLowerCase();
for (int i = 0; i <= lower.length() - shingleSize; i++) {
shingles.add(lower.substring(i, i + shingleSize));
}
return shingles;
}
/**
* Converts a shingle set to a boolean presence vector over the shared vocabulary.
*
* @param shingles shingle set for this sentence
* @param vocabList ordered list of all vocabulary shingles
* @return boolean[] where true = shingle present
*/
private boolean[] toVector(Set<String> shingles, List<String> vocabList) {
boolean[] vector = new boolean[vocabList.size()];
for (int i = 0; i < vocabList.size(); i++) {
vector[i] = shingles.contains(vocabList.get(i));
}
return vector;
}
// -----------------------------------------------------------------------
// Utility
// -----------------------------------------------------------------------
private String sanitiseTsv(String s) {
if (s == null) return "";
return s.replace("\t", " ").replace("\n", " ").replace("\r", "");
}
/** Returns an unmodifiable view of all detected duplicate pairs. */
public List<DuplicatePair> getDuplicatePairs() {
return Collections.unmodifiableList(duplicatePairs);
}
/** Returns the number of corpus sentences indexed. */
public int getCorpusSize() {
return corpusIndex.size();
}
// -----------------------------------------------------------------------
// Inner data classes
// -----------------------------------------------------------------------
/**
* Uniquely identifies a sentence by its source file name and line number.
*/
public static class SentenceKey {
public final String fileName;
public final int lineNumber;
public SentenceKey(String fileName, int lineNumber) {
this.fileName = fileName;
this.lineNumber = lineNumber;
}
@Override
public boolean equals(Object o) {
if (!(o instanceof SentenceKey)) return false;
SentenceKey other = (SentenceKey) o;
return lineNumber == other.lineNumber
&& fileName.equals(other.fileName);
}
@Override
public int hashCode() {
return 31 * fileName.hashCode() + lineNumber;
}
@Override
public String toString() {
return fileName + ":" + lineNumber;
}
}
/**
* Holds the raw text and MinHash signature for an indexed corpus sentence.
*/
private static class IndexedSentence {
final String text;
final int[] signature;
IndexedSentence(String text, int[] signature) {
this.text = text;
this.signature = signature;
}
}
/**
* Represents a detected near-duplicate pair between a new sentence
* and a corpus sentence.
*/
public static class DuplicatePair {
public final SentenceKey newKey;
public final String newText;
public final SentenceKey corpusKey;
public final String corpusText;
public final double similarity;
public DuplicatePair(SentenceKey newKey, String newText,
SentenceKey corpusKey, String corpusText,
double similarity) {
this.newKey = newKey;
this.newText = newText;
this.corpusKey = corpusKey;
this.corpusText = corpusText;
this.similarity = similarity;
}
@Override
public String toString() {
return String.format("[%.2f] %s ↔ %s", similarity, newKey, corpusKey);
}
}
}
|