Examples of edu.umd.cloud9.io.map.HMapSIW

edu.umd.cloud9.io.pair.PairOfFloats
Writable representing a map where keys are Strings and values are ints. This class is specialized for String objects to avoid the overhead that comes with wrapping Strings inside Text objects. @author Jimmy Lin

    String label;
    long time = System.currentTimeMillis();


    for (int i = 0; i < transVectors.size(); i++) {
      HMapSFW transVector = transVectors.get(i);
      HMapSIW fTfMap = fTfs.get(i);
      String fSent = fSents.get(i);
      for (int j = 0; j < eVectors.size(); j++) {
        HMapSFW eVector = eVectors.get(j);
        HMapSIW eTfMap = eTfs.get(j);
        String eSent = eSents.get(j);
        if (parallelPairs.get(i) == j) {
          label = "parallel";
        } else {
          label = "non_parallel";

View Full Code Here

    String VOCABDIR = "data/vocab";    // /Users/ferhanture/Documents/workspace/ivory-github/Ivory/data/vocab
    String TOKENDIR = "data/tokenizer";   // "/Users/ferhanture/Documents/workspace/ivory-github/Ivory/data/tokenizer
    String DATADIR = "/fs/clip-qa/ferhan/cl-pwsim/pwsim-experiments-2013";    // /Users/ferhanture/edu/research_archive/data/de-en/eu-nc-wmt08
    
    BitextClassifierUtils dt = new BitextClassifierUtils();
    numSentencesPerDocE = new HMapSIW();
    numSentencesPerDocF = new HMapSIW();
    FileSystem localFs = FileSystem.getLocal(new Configuration());
    eVocabSrc = HadoopAlign.loadVocab(new Path(VOCABDIR+"/vocab.en-de.en"), localFs);
    eVocabTrg = HadoopAlign.loadVocab(new Path(VOCABDIR+"/vocab.de-en.en"), localFs);
    fVocabSrc = HadoopAlign.loadVocab(new Path(VOCABDIR+"/vocab.de-en.de"), localFs);
    fVocabTrg = HadoopAlign.loadVocab(new Path(VOCABDIR+"/vocab.en-de.de"), localFs);

View Full Code Here

      return null;
    }
    PriorityQueue<PairOfFloatInt> eS = f2eProbs.get(f).getTranslationsWithProbs(lexProbThreshold);


    if (!eS.isEmpty()) {
      PairOfFloatInt entry = eS.poll();
      int e = entry.getRightElement();
      String eTerm = eVocab_f2e.get(e);
      return eTerm;
    }
    return token;
  }

View Full Code Here


    float sumProbEF = 0;
    int numTrans = 0;
    //tf(e) = sum_f{tf(f)*prob(e|f)}
    while (numTrans < numTransPerToken && !eS.isEmpty()) {
      PairOfFloatInt entry = eS.poll();
      float probEF = entry.getLeftElement();
      int e = entry.getRightElement();
      String eTerm = eVocab_f2e.get(e);


      //      LOG.info("Pr("+eTerm+"|"+token+")="+probEF);


      if (probEF > 0 && e > 0 && !docLangTokenizer.isStopWord(eTerm) && (translateOnly == null || !translateOnly.equals("indri") || indriPuncPattern.matcher(eTerm).matches()) && (pairsInSCFG == null || pairsInSCFG.contains(new PairOfStrings(token,eTerm)))) {

View Full Code Here

            curIndex = prevIndex;    // revert curIndex value since we're skipping this one
            skipTerm = true;
            continue;
          }
          logger.debug("Processing: "+srcTerm+" with index: "+curIndex);      
          topTrans.add(new PairOfFloatString(prob, trgTerm));
          sumOfProbs += prob;
          logger.debug("Added to queue: "+trgTerm+" with prob: "+prob+" (sum: "+sumOfProbs+")");      
        }else if(!earlyTerminate && !skipTerm && !delims.contains(srcTerm)){  //continue adding translation term,prob pairs (except if early termination is ON)
          topTrans.add(new PairOfFloatString(prob, trgTerm));
          sumOfProbs += prob;
          logger.debug("Added to queue: "+trgTerm+" with prob: "+prob+" (sum: "+sumOfProbs+")");      


          // keep top numTrans translations
          if(topTrans.size() > numTrans){
            PairOfFloatString pair = topTrans.pollFirst();
            float removedProb = pair.getLeftElement();
            sumOfProbs -= removedProb;
            logger.debug("Removed from queue: "+pair.getRightElement()+" (sum: "+sumOfProbs+")");      
          }
        }else{
          logger.debug("Skipped line: "+line);
        }
      }

View Full Code Here

          continue;
        }
        prob = ttable.get(srcIndex, trgIndex);
        logger.debug("Found: " + trgTerm + " with " + prob);


        topTrans.add(new PairOfFloatString(prob, trgTerm));
        // keep top numTrans translations
        if (topTrans.size() > numTrans) {
          float removedProb = topTrans.pollFirst().getLeftElement();
          sumOfProbs -= removedProb;
        }

View Full Code Here

    List<Integer> sortedIndices = new ArrayList<Integer>();
    HMapIF index2ProbMap = new HMapIF();


    float sumOfProbs = 0.0f;    //only extract the top K<15 if the mass prob. exceeds MAX_probThreshold
    while(!topTrans.isEmpty() && sumOfProbs < cumProbThreshold){
      PairOfFloatString e = topTrans.pollLast();
      String term = e.getRightElement();
      float pr = e.getLeftElement()/cumProb;    // normalize
      logger.debug(term+"-->"+pr);
      int trgIndex = trgVocab.addOrGet(term);
      sumOfProbs += e.getLeftElement();         // keep track of unnormalized cumulative prob for determining cutoff
      sortedIndices.add(trgIndex);
      index2ProbMap.put(trgIndex, pr);
    }


    // to enable faster access with binary search, we sort entries by vocabulary index.

View Full Code Here

          int e2 = eVocabTrg.get(eTerm);         


          float prob2 = f2e_Probs.get(f2, e2);
          float prob = prob1*prob2;
          sumOfProbs += prob;
          topTrans.add(new PairOfFloatString(prob, fTerm));
        }
        logger.info("Adding "+eTerm);
        addToTable(e1, topTrans, sumOfProbs, table, fVocabTrg, 1.0f, stats);      
      }
      logger.info(stats);

View Full Code Here

    if (featSet > 2) {
      // uppercase token matching features : find uppercased tokens that exactly appear on both sides
      // lack of this evidence does not imply anything, but its existence might indicate parallel
//      fSentence.replaceAll("([',:;.?%!])", " $1 ");
//      eSentence.replaceAll("([',:;.?%!])", " $1 ");
      PairOfFloats pair = getUppercaseRatio(fTokenizer.processContent(fSentence), eTokenizer.processContent(eSentence));
      features.add("uppercaseratio1=" + pair.getLeftElement() );
      features.add("uppercaseratio2=" + pair.getRightElement() );
    }


    if (featSet > 3) {
      // future work = count number of single/double letter words in src and trg side

View Full Code Here

    // now, read tokens in first sentence and keep track of sequences of uppercased tokens in buffer
    HashSet<String> upperCaseMap1 = getUppercaseParts(tokens1);
    HashSet<String> upperCaseMap2 = getUppercaseParts(tokens2);
    float cntUpperRatio1 = getRatio(upperCaseMap1, upperCaseMap2);
    float cntUpperRatio2 = getRatio(upperCaseMap2, upperCaseMap1);
    PairOfFloats result = new PairOfFloats(cntUpperRatio1, cntUpperRatio2);
    return result;
  }

View Full Code Here

0 1 2 3 4 5 6 7 8 9

TOP

Related Classes of edu.umd.cloud9.io.map.HMapSIW

bak.pcj.IntIterator

edu.umd.cloud9.example.bfs.BfsNodeTest

edu.umd.cloud9.example.pagerank.PageRankNodeTest

edu.umd.cloud9.io.array.ArrayListOfIntsWritable

edu.umd.cloud9.io.fastutil.Int2FloatOpenHashMapWritableTest

edu.umd.cloud9.io.fastutil.Int2IntOpenHashMapWritableTest

edu.umd.cloud9.io.fastutil.String2FloatOpenHashMapWritableTest

edu.umd.cloud9.io.fastutil.String2IntOpenHashMapWritableTest

edu.umd.cloud9.io.map.HMapIDW

edu.umd.cloud9.io.map.HMapIDWTest

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.