/**
   * Sets boost of termClaimsDescriptionAbstractTitles. boost = weight = factor(tf*idf)
   *
   * @param vecsTerms
   * @param currentField
   * @param factor - adjustment factor ( ex. alpha or beta )
   * @param decayFactor
   * @return
   * @throws java.io.IOException
   */
  public Map<String, TermQuery> setBoost(
      Map<TermFreqVector, String> vecsTerms, String currentField, float factor, float decayFactor)
      throws IOException {
    Map<String, TermQuery> terms = new HashMap<>();
    // setBoost for each of the terms of each of the docs
    int i = 0;
    float norm = (float) 1 / vecsTerms.size();
    //        System.out.println("--------------------------");
    for (Map.Entry<TermFreqVector, String> e : vecsTerms.entrySet()) {
      // Increase decay
      String field = e.getValue();
      TermFreqVector docTerms = e.getKey();
      float decay = decayFactor * i;
      // Populate terms: with TermQuries and set boost
      for (String termTxt : docTerms.getTerms()) {
        // Create Term
        Term term = new Term(currentField, termTxt);
        // Calculate weight
        float tf = docTerms.getFreq(termTxt);
        //                float idf = ir.docFreq(termTitle);
        int docs;
        float idf;
        if (sourceField.equals(PatentQuery.all)) {
          docs = ir.getDocCount(field);
          idf = (float) Math.log10((double) docs / (ir.docFreq(new Term(field, termTxt)) + 1));
        } else {
          docs = ir.getDocCount(sourceField);
          idf =
              (float) Math.log10((double) docs / (ir.docFreq(new Term(sourceField, termTxt)) + 1));
        }
        float weight = tf * idf;

        //                System.out.println(term.text() + " -> tf= " + tf + " idf= " + idf + "
        // tfidf= " + weight);
        // Adjust weight by decay factor
        weight = weight - (weight * decay);
        // Create TermQuery and add it to the collection
        TermQuery termQuery = new TermQuery(term);
        // Calculate and set boost
        float boost;
        if (vecsTerms.size() == 1) {
          boost = factor * tf;
        } else {
          boost = factor;
        }

        if (boost != 0) {
          termQuery.setBoost(boost * norm);
          if (terms.containsKey(termTxt)) {
            TermQuery tq = terms.get(termTxt);
            tq.setBoost(tq.getBoost() + termQuery.getBoost());
          } else {
            terms.put(termTxt, termQuery);
          }
        }
      }
      i++;
    }
    return terms;
  }
Пример #2
0
  public static void main(String[] args) throws IOException {

    IndexReader reader = null;

    /*
     *  Opening the index first simplifies the processing of the
     *  rest of the command line arguments.
     */
    for (int i = 0; i < args.length; i++) {
      if (("-index".equals(args[i])) && ((i + 1) < args.length)) {
        reader = DirectoryReader.open(FSDirectory.open(new File(args[i + 1])));

        if (reader == null) {
          System.err.println("Error:  Can't open index " + args[i + 1]);
          System.exit(1);
        }
        ;

        break;
      }
      ;
    }
    ;

    if (reader == null) {
      System.err.println(usage);
      System.exit(1);
    }
    ;

    /*
     *  Process the command line arguments sequentially.
     */
    for (int i = 0; i < args.length; i++) {

      if ("-index".equals(args[i])) {

        /*
         *  Handled in the previous loop, so just skip the argument.
         */
        i++;

      } else if ("-list-edocid".equals(args[i])) {

        System.out.println("-list-edocid:");

        if ((i + 1) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        Document d = reader.document(Integer.parseInt(args[i + 1]));

        System.out.println(
            "Internal docid --> External docid: " + args[i + 1] + " --> " + d.get("externalId"));

        i += 1;
      } else if ("-list-docids".equals(args[i])) {

        System.out.println("-list-docids:");

        for (int j = 0; j < reader.numDocs(); j++) {
          Document d = reader.document(j);
          System.out.println("Internal --> external docid: " + j + " --> " + d.get("externalId"));
        }
        ;

      } else if ("-list-fields".equals(args[i])) {

        Fields fields = MultiFields.getFields(reader);

        System.out.print("\nNumber of fields:  ");

        if (fields == null) System.out.println("0");
        else {
          System.out.println(fields.size());

          Iterator<String> is = fields.iterator();

          while (is.hasNext()) {
            System.out.println("\t" + is.next());
          }
          ;
        }
        ;

      } else if ("-list-postings".equals(args[i])) {

        if ((i + 2) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        listPostings(reader, args[i + 1], args[i + 2], Integer.MAX_VALUE);
        i += 2;

      } else if ("-list-postings-sample".equals(args[i])) {

        if ((i + 2) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        listPostings(reader, args[i + 1], args[i + 2], 5);
        i += 2;

      } else if ("-list-stats".equals(args[i])) {

        System.out.println("Corpus statistics:");
        System.out.println("\tnumdocs\t\t" + reader.numDocs());
        System.out.println(
            "\turl:\t"
                + "\tnumdocs="
                + reader.getDocCount("url")
                + "\tsumTotalTF="
                + reader.getSumTotalTermFreq("url")
                + "\tavglen="
                + reader.getSumTotalTermFreq("url") / (float) reader.getDocCount("url"));

        System.out.println(
            "\tkeywords:"
                + "\tnumdocs="
                + reader.getDocCount("keywords")
                + "\tsumTotalTF="
                + reader.getSumTotalTermFreq("keywords")
                + "\tavglen="
                + reader.getSumTotalTermFreq("keywords") / (float) reader.getDocCount("keywords"));

        System.out.println(
            "\ttitle:\t"
                + "\tnumdocs="
                + reader.getDocCount("title")
                + "\tsumTotalTF="
                + reader.getSumTotalTermFreq("title")
                + "\tavglen="
                + reader.getSumTotalTermFreq("title") / (float) reader.getDocCount("title"));

        System.out.println(
            "\tbody:\t"
                + "\tnumdocs="
                + reader.getDocCount("body")
                + "\tsumTotalTF="
                + reader.getSumTotalTermFreq("body")
                + "\tavglen="
                + reader.getSumTotalTermFreq("body") / (float) reader.getDocCount("body"));

        System.out.println(
            "\tinlink:\t"
                + "\tnumdocs="
                + reader.getDocCount("inlink")
                + "\tsumTotalTF="
                + reader.getSumTotalTermFreq("inlink")
                + "\tavglen="
                + reader.getSumTotalTermFreq("inlink") / (float) reader.getDocCount("inlink"));

      } else if ("-list-terms".equals(args[i])) {

        if ((i + 1) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        listTermDictionary(reader, args[i + 1]);
        i += 1;

      } else if ("-list-termvector".equals(args[i])) {

        if ((i + 1) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        listTermVectors(reader, args[i + 1]);
        i += 1;

      } else if ("-list-termvector-field".equals(args[i])) {

        if ((i + 2) >= args.length) {
          System.out.println(usage);
          break;
        }
        ;

        listTermVectorField(reader, args[i + 1], args[i + 2]);
        i += 2;

      } else System.err.println("\nWarning:  Unknown argument " + args[i] + " ignored.");
    }
    ;

    /*
     *  Close the index and exit gracefully.
     */
    reader.close();
  }