コード例 #1
0
 private boolean containsMostlyLetters(BxLine line) {
   double letterCount = 0;
   for (char ch : line.toText().toCharArray()) {
     if (Character.isLetter(ch)) letterCount++;
   }
   return 2 * letterCount > line.toText().length();
 }
コード例 #2
0
 private boolean shouldBeRemoved(
     BxLine line,
     Population heightPopulation,
     Population fontPopulation,
     Population distancePopulation,
     Population indentationPopulation) {
   if (line.getMostPopularFontName() == null) {
     return true;
   }
   if (heightPopulation.getZScore(line.getHeight()) < candMinHeightZScore) {
     return true;
   }
   if (looksLikeEquation(line)) {
     return true;
   }
   if (looksLikeFigure(line)) {
     return true;
   }
   if (looksLikeTable(line)) {
     return true;
   }
   if (!containsMostlyLetters(line)) {
     return true;
   }
   if (!containsWord(line)) {
     return true;
   }
   if (startsWithLargeNumber(line)) {
     return true;
   }
   if (heightPopulation.getZScore(line.getHeight()) < outlHeightZScore
       && Math.abs(fontPopulation.getZScore(getFontIndex(line))) < outlFontZScore
       && (!line.hasPrev()
           || distancePopulation.getZScore(line.getY() - line.getPrev().getY())
               < outlDistanceZScore)
       && Math.abs(indentationPopulation.getZScore(line.getX())) < outlIndentZScore) {
     return true;
   }
   int i = 0;
   BxLine actLine = line;
   while (actLine.hasNext()) {
     actLine = actLine.getNext();
     if (actLine.toText().matches("[A-Z].*")) {
       break;
     }
     if (i++ == maxHeaderLineCount) {
       return true;
     }
   }
   return false;
 }
コード例 #3
0
ファイル: ContentCleaner.java プロジェクト: matfed/CERMINE
  public void cleanupContent(BxDocContentStructure contentStructure) {
    for (BxDocContentPart contentPart : contentStructure.getParts()) {
      List<BxLine> headerLines = contentPart.getHeaderLines();
      StringBuilder sb = new StringBuilder();
      for (BxLine headerLine : headerLines) {
        String lineText = headerLine.toText();
        if (lineText.endsWith("-")) {
          lineText = lineText.substring(0, lineText.length() - 1);
          if (lineText.lastIndexOf(' ') < 0) {
            sb.append(lineText);
          } else {
            sb.append(lineText.substring(0, lineText.lastIndexOf(' ')));
            sb.append(" ");
            sb.append(lineText.substring(lineText.lastIndexOf(' ') + 1));
          }
        } else {
          sb.append(lineText);
          sb.append(" ");
        }
      }
      contentPart.setCleanHeaderText(cleanLigatures(sb.toString().trim()));

      List<BxLine> contentLines = contentPart.getContentLines();
      List<String> contentTexts = new ArrayList<String>();

      double maxLen = Double.NEGATIVE_INFINITY;
      for (BxLine line : contentLines) {
        if (line.getWidth() > maxLen) {
          maxLen = line.getWidth();
        }
      }

      String contentText = "";
      for (BxLine line : contentLines) {
        int score = 0;
        BxLine prev = line.getPrev();
        BxLine next = line.getNext();
        if (line.toText().matches("^[A-Z].*$")) {
          score++;
        }
        if (prev != null) {
          if (line.getX() > prev.getX()
              && line.getX() - prev.getX() < paragraphLineIndentMultiplier * maxLen
              && line.getX() - prev.getX() > minParagraphIndent) {
            score++;
          }
          if (prev.getWidth() < lastParagraphLineLengthMult * maxLen) {
            score++;
          }
          if (prev.toText().endsWith(".")) {
            score++;
          }
        }
        if (next != null
            && line.getX() > next.getX()
            && line.getX() - next.getX() < paragraphLineIndentMultiplier * maxLen
            && line.getX() - next.getX() > minParagraphIndent) {
          score++;
        }

        if (score >= firstParagraphLineMinScore) {
          if (!contentText.isEmpty()) {
            contentTexts.add(cleanLigatures(contentText.trim()));
          }
          contentText = "";
        }

        String lineText = line.toText();
        if (lineText.endsWith("-")) {
          lineText = lineText.substring(0, lineText.length() - 1);
          if (lineText.lastIndexOf(' ') < 0) {
            contentText += lineText;
          } else {
            contentText += lineText.substring(0, lineText.lastIndexOf(' '));
            contentText += "\n";
            contentText += lineText.substring(lineText.lastIndexOf(' ') + 1);
          }
        } else {
          contentText += lineText;
          contentText += "\n";
        }
      }
      if (!contentText.isEmpty()) {
        contentTexts.add(cleanLigatures(contentText.trim()));
      }

      contentPart.setCleanContentTexts(contentTexts);
    }
  }
コード例 #4
0
 private boolean startsWithLargeNumber(BxLine line) {
   return line.toText().matches("[0-9][0-9].*");
 }
コード例 #5
0
 private boolean containsWord(BxLine line) {
   return line.toText().toLowerCase().matches(".*[a-z][a-z][a-z][a-z].*");
 }
コード例 #6
0
 private boolean looksLikeTable(BxLine line) {
   return line.toText().toLowerCase().matches("table .*");
 }
コード例 #7
0
 private boolean looksLikeFigure(BxLine line) {
   return line.toText().toLowerCase().matches("fig\\.? .*")
       || line.toText().toLowerCase().matches("figure .*");
 }
コード例 #8
0
 private boolean looksLikeEquation(BxLine line) {
   return line.toText().contains("=");
 }
コード例 #9
0
 private boolean looksLikeHeader(BxLine line) {
   String text = line.toText();
   return text.matches("^[A-Z].*")
       || text.matches("^[1-9].*[a-zA-Z].*")
       || text.matches("^[a-h]\\).*[a-zA-Z].*");
 }
コード例 #10
0
 @Override
 public double calculateFeatureValue(BxLine line, BxPage page) {
   return (line.toText().matches("^[a-z]\\) [A-Z].*$")) ? 1 : 0;
 }