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();
}
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;
}
private boolean startsWithLargeNumber(BxLine line) {
return line.toText().matches("[0-9][0-9].*");
}
private boolean containsWord(BxLine line) {
return line.toText().toLowerCase().matches(".*[a-z][a-z][a-z][a-z].*");
}
private boolean looksLikeTable(BxLine line) {
return line.toText().toLowerCase().matches("table .*");
}
private boolean looksLikeFigure(BxLine line) {
return line.toText().toLowerCase().matches("fig\\.? .*")
|| line.toText().toLowerCase().matches("figure .*");
}
private boolean looksLikeEquation(BxLine line) {
return line.toText().contains("=");
}
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].*");
}
