/**
* Create the set of training and evaluation sets from the annotated examples with extraction of
* citations in the patent description body.
*
* @param rank rank associated to the set for n-fold data generation
* @param type type of data to be created, 0 is training data, 1 is evaluation data
*/
public void createDataSet(
String setName, String rank, String corpusPath, String outputPath, int type) {
int nbFiles = 0;
int nbNPLRef = 0;
int nbPatentRef = 0;
int maxRef = 0;
try {
// PATENT REF. textual data
// we use a SAX parser on the patent XML files
MarecSaxParser sax = new MarecSaxParser();
sax.patentReferences = true;
sax.nplReferences = false;
int srCitations = 0;
int previousSrCitations = 0;
int withSR = 0;
List<OffsetPosition> journalsPositions = null;
List<OffsetPosition> abbrevJournalsPositions = null;
List<OffsetPosition> conferencesPositions = null;
List<OffsetPosition> publishersPositions = null;
if (type == 0) {
// training set
sax.setN(trainWindow);
} else {
// for the test set we enlarge the focus window to include all the document.
sax.setN(-1);
}
// get a factory
/*SAXParserFactory spf = SAXParserFactory.newInstance();
spf.setValidating(false);
spf.setFeature("http://xml.org/sax/features/namespaces", false);
spf.setFeature("http://xml.org/sax/features/validation", false);
LinkedList<File> fileList = new LinkedList<File>();
if (setName == null) {
fileList.add(new File(corpusPath));
} else if (rank == null) {
fileList.add(new File(corpusPath));
} else {
// n-fold evaluation
fileList.add(new File(corpusPath + File.separator + setName + "ing" + rank + File.separator));
}
Writer writer = null;
if ((setName == null) || (setName.length() == 0)) {
writer = new OutputStreamWriter(new FileOutputStream(
new File(outputPath + "/patent.train"), false), "UTF-8");
} else if (rank == null) {
writer = new OutputStreamWriter(new FileOutputStream(
new File(outputPath + "/patent." + setName), false), "UTF-8");
} else {
writer = new OutputStreamWriter(new FileOutputStream(
new File(outputPath + setName + "ing" + rank + "/patent." + setName), false), "UTF-8");
}
while (fileList.size() > 0) {
File file = fileList.removeFirst();
if (file.isDirectory()) {
for (File subFile : file.listFiles())
fileList.addLast(subFile);
} else {
if (file.getName().endsWith(".xml")) {
nbFiles++;
System.out.println(file.getAbsolutePath());
try {
//get a new instance of parser
SAXParser p = spf.newSAXParser();
FileInputStream in = new FileInputStream(file);
sax.setFileName(file.getName());
p.parse(in, sax);
//writer1.write("\n");
nbPatentRef += sax.getNbPatentRef();
if (sax.citations != null) {
if (sax.citations.size() > previousSrCitations) {
previousSrCitations = sax.citations.size();
withSR++;
}
}
journalsPositions = sax.journalsPositions;
abbrevJournalsPositions = sax.abbrevJournalsPositions;
conferencesPositions = sax.conferencesPositions;
publishersPositions = sax.publishersPositions;
if (sax.accumulatedText != null) {
String text = sax.accumulatedText.toString();
if (text.trim().length() > 0) {
// add features for the patent tokens
addFeatures(text,
writer,
journalsPositions,
abbrevJournalsPositions,
conferencesPositions,
publishersPositions);
writer.write("\n \n");
}
}
} catch (Exception e) {
throw new GrobidException("An exception occured while running Grobid.", e);
}
}
}
}*/
// NPL REF. textual data
/*sax = new MarecSaxParser();
sax.patentReferences = false;
sax.nplReferences = true;
if (type == 0) {
// training set
sax.setN(trainWindow);
}
else {
// for the test set we enlarge the focus window to include all the document.
sax.setN(-1);
}
// get a factory
spf = SAXParserFactory.newInstance();
spf.setValidating(false);
spf.setFeature("http://xml.org/sax/features/namespaces", false);
spf.setFeature("http://xml.org/sax/features/validation", false);
fileList = new LinkedList<File>();
if (setName == null) {
fileList.add(new File(corpusPath));
} else if (rank == null) {
fileList.add(new File(corpusPath));
} else {
fileList.add(new File(corpusPath + File.separator + setName + "ing" + rank + File.separator));
}
if ((setName == null) || (setName.length() == 0)) {
writer = new OutputStreamWriter(new FileOutputStream(
new File(outputPath + "/npl.train"), false), "UTF-8");
} else if (rank == null) {
writer = new OutputStreamWriter(new FileOutputStream(
new File(outputPath + "/npl." + setName), false), "UTF-8");
} else {
writer = new OutputStreamWriter(new FileOutputStream(
new File(outputPath + File.separator + setName + "ing" + rank + File.separator +
"npl." + setName), false), "UTF-8");
}
while (fileList.size() > 0) {
File file = fileList.removeFirst();
if (file.isDirectory()) {
for (File subFile : file.listFiles())
fileList.addLast(subFile);
} else {
if (file.getName().endsWith(".xml")) {
//nbFiles++;
//String text = Files.readFromFile(file,"UTF-8");
try {
//get a new instance of parser
SAXParser p = spf.newSAXParser();
FileInputStream in = new FileInputStream(file);
sax.setFileName(file.toString());
p.parse(in, sax);
//writer2.write("\n");
nbNPLRef += sax.getNbNPLRef();
if (sax.nbAllRef > maxRef) {
maxRef = sax.nbAllRef;
}
if (sax.citations != null) {
if (sax.citations.size() > previousSrCitations) {
previousSrCitations = sax.citations.size();
withSR++;
}
}
journalsPositions = sax.journalsPositions;
abbrevJournalsPositions = sax.abbrevJournalsPositions;
conferencesPositions = sax.conferencesPositions;
publishersPositions = sax.publishersPositions;
//totalLength += sax.totalLength;
if (sax.accumulatedText != null) {
String text = sax.accumulatedText.toString();
// add features for NPL
addFeatures(text,
writer,
journalsPositions,
abbrevJournalsPositions,
conferencesPositions,
publishersPositions);
writer.write("\n");
}
} catch (Exception e) {
throw new GrobidException("An exception occured while running Grobid.", e);
}
}
}
}
if (sax.citations != null)
srCitations += sax.citations.size();*/
// Patent + NPL REF. textual data (the "all" model)
sax = new MarecSaxParser();
sax.patentReferences = true;
sax.nplReferences = true;
if (type == 0) {
// training set
sax.setN(trainWindow);
} else {
// for the test set we enlarge the focus window to include all the document.
sax.setN(-1);
}
// get a factory
SAXParserFactory spf = SAXParserFactory.newInstance();
spf.setValidating(false);
spf.setFeature("http://xml.org/sax/features/namespaces", false);
spf.setFeature("http://xml.org/sax/features/validation", false);
LinkedList<File> fileList = new LinkedList<File>();
if (setName == null) {
fileList.add(new File(corpusPath));
} else if (rank == null) {
fileList.add(new File(corpusPath));
} else {
fileList.add(
new File(corpusPath + File.separator + setName + "ing" + rank + File.separator));
}
Writer writer = null;
if ((setName == null) || (setName.length() == 0)) {
writer =
new OutputStreamWriter(
new FileOutputStream(new File(outputPath + File.separator + "all.train"), false),
"UTF-8");
} else if (rank == null) {
writer =
new OutputStreamWriter(
new FileOutputStream(
new File(outputPath + File.separator + "all." + setName), false),
"UTF-8");
} else {
writer =
new OutputStreamWriter(
new FileOutputStream(
new File(
outputPath
+ File.separator
+ setName
+ "ing"
+ rank
+ File.separator
+ "all."
+ setName),
false),
"UTF-8");
}
// int totalLength = 0;
while (fileList.size() > 0) {
File file = fileList.removeFirst();
if (file.isDirectory()) {
for (File subFile : file.listFiles()) {
fileList.addLast(subFile);
}
} else {
if (file.getName().endsWith(".xml")) {
nbFiles++;
try {
// get a new instance of parser
SAXParser p = spf.newSAXParser();
FileInputStream in = new FileInputStream(file);
sax.setFileName(file.toString());
p.parse(in, sax);
// writer3.write("\n");
nbNPLRef += sax.getNbNPLRef();
nbPatentRef += sax.getNbPatentRef();
if (sax.nbAllRef > maxRef) {
maxRef = sax.nbAllRef;
}
if (sax.citations != null) {
if (sax.citations.size() > previousSrCitations) {
previousSrCitations = sax.citations.size();
withSR++;
}
}
journalsPositions = sax.journalsPositions;
abbrevJournalsPositions = sax.abbrevJournalsPositions;
conferencesPositions = sax.conferencesPositions;
publishersPositions = sax.publishersPositions;
// totalLength += sax.totalLength;
if (sax.accumulatedText != null) {
String text = sax.accumulatedText.toString();
// add features for patent+NPL
addFeatures(
text,
writer,
journalsPositions,
abbrevJournalsPositions,
conferencesPositions,
publishersPositions);
writer.write("\n");
}
} catch (Exception e) {
throw new GrobidException("An exception occured while running Grobid.", e);
}
}
}
}
if (sax.citations != null) {
srCitations += sax.citations.size();
}
if (setName != null) {
System.out.println(setName + "ing on " + nbFiles + " files");
} else {
System.out.println("training on " + nbFiles + " files");
}
// System.out.println("Number of file with search report: " + withSR);
System.out.println("Number of references: " + (nbNPLRef + nbPatentRef));
System.out.println("Number of patent references: " + nbPatentRef);
System.out.println("Number of NPL references: " + nbNPLRef);
// System.out.println("Number of search report citations: " + srCitations);
System.out.println(
"Average number of references: "
+ TextUtilities.formatTwoDecimals((double) (nbNPLRef + nbPatentRef) / nbFiles));
System.out.println("Max number of references in file: " + maxRef);
/*if ((setName == null) || (setName.length() == 0)) {
System.out.println("patent data set under: " + outputPath + "/patent.train");
} else {
System.out.println("patent data set under: " + outputPath + "/patent." + setName);
}
if ((setName == null) || (setName.length() == 0)) {
System.out.println("npl data set under: " + outputPath + "/npl.train");
} else {
System.out.println("npl data set under: " + outputPath + "/npl." + setName);
}*/
if ((setName == null) || (setName.length() == 0)) {
System.out.println("common data set under: " + outputPath + "/all.train");
} else {
System.out.println("common data set under: " + outputPath + "/all." + setName);
}
} catch (Exception e) {
throw new GrobidException("An exception occurred while running Grobid.", e);
}
}
/**
* Extract results from a labelled full text in the training format without any string
* modification.
*
* @param result reult
* @param tokenizations toks
* @return extraction
*/
private StringBuffer trainingExtraction(String result, ArrayList<String> tokenizations) {
// this is the main buffer for the whole full text
StringBuffer buffer = new StringBuffer();
try {
StringTokenizer st = new StringTokenizer(result, "\n");
String s1 = null;
String s2 = null;
String lastTag = null;
// current token position
int p = 0;
boolean start = true;
boolean openFigure = false;
boolean headFigure = false;
boolean descFigure = false;
boolean tableBlock = false;
while (st.hasMoreTokens()) {
boolean addSpace = false;
String tok = st.nextToken().trim();
if (tok.length() == 0) {
continue;
}
StringTokenizer stt = new StringTokenizer(tok, " \t");
ArrayList<String> localFeatures = new ArrayList<String>();
int i = 0;
boolean newLine = false;
int ll = stt.countTokens();
while (stt.hasMoreTokens()) {
String s = stt.nextToken().trim();
if (i == 0) {
s2 = TextUtilities.HTMLEncode(s); // lexical token
boolean strop = false;
while ((!strop) && (p < tokenizations.size())) {
String tokOriginal = tokenizations.get(p);
if (tokOriginal.equals(" ")) {
addSpace = true;
} else if (tokOriginal.equals(s)) {
strop = true;
}
p++;
}
} else if (i == ll - 1) {
s1 = s; // current tag
} else {
if (s.equals("LINESTART")) newLine = true;
localFeatures.add(s);
}
i++;
}
if (newLine && !start) {
buffer.append("<lb/>");
}
String lastTag0 = null;
if (lastTag != null) {
if (lastTag.startsWith("I-")) {
lastTag0 = lastTag.substring(2, lastTag.length());
} else {
lastTag0 = lastTag;
}
}
String currentTag0 = null;
if (s1 != null) {
if (s1.startsWith("I-")) {
currentTag0 = s1.substring(2, s1.length());
} else {
currentTag0 = s1;
}
}
boolean closeParagraph = false;
if (lastTag != null) {
closeParagraph = testClosingTag(buffer, currentTag0, lastTag0, s1);
}
boolean output = false;
if (!currentTag0.equals("<table>")
&& !currentTag0.equals("<trash>")
&& !currentTag0.equals("<figure_head>")
&& !currentTag0.equals("<label>")) {
if (openFigure) {
buffer.append("\n\t\t\t</figure>\n\n");
}
openFigure = false;
headFigure = false;
descFigure = false;
tableBlock = false;
}
output = writeField(buffer, s1, lastTag0, s2, "<header>", "<front>", addSpace, 3);
if (!output) {
output =
writeField(buffer, s1, lastTag0, s2, "<other>", "<note type=\"other\">", addSpace, 3);
}
// for paragraph we must distinguish starting and closing tags
if (!output) {
if (closeParagraph) {
output = writeFieldBeginEnd(buffer, s1, "", s2, "<paragraph>", "<p>", addSpace, 3);
} else {
output = writeFieldBeginEnd(buffer, s1, lastTag, s2, "<paragraph>", "<p>", addSpace, 3);
}
}
if (!output) {
output =
writeField(
buffer,
s1,
lastTag0,
s2,
"<page_header>",
"<note place=\"headnote\">",
addSpace,
3);
}
if (!output) {
output =
writeField(
buffer,
s1,
lastTag0,
s2,
"<page_footnote>",
"<note place=\"footnote\">",
addSpace,
3);
}
if (!output) {
output = writeField(buffer, s1, lastTag0, s2, "<page>", "<page>", addSpace, 3);
}
if (!output) {
output =
writeFieldBeginEnd(buffer, s1, lastTag0, s2, "<reference>", "<bibl>", addSpace, 3);
}
if (!output) {
if (closeParagraph) {
output = writeField(buffer, s1, "", s2, "<reference_marker>", "<label>", addSpace, 3);
} else
output =
writeField(buffer, s1, lastTag0, s2, "<reference_marker>", "<label>", addSpace, 3);
}
if (!output) {
output =
writeField(
buffer,
s1,
lastTag0,
s2,
"<citation_marker>",
"<ref type=\"biblio\">",
addSpace,
3);
}
if (!output) {
output = writeField(buffer, s1, lastTag0, s2, "<section>", "<head>", addSpace, 3);
}
if (!output) {
output = writeField(buffer, s1, lastTag0, s2, "<subsection>", "<head>", addSpace, 3);
}
if (!output) {
if (openFigure) {
output = writeField(buffer, s1, lastTag0, s2, "<trash>", "<trash>", addSpace, 4);
} else {
// output = writeField(buffer, s1, lastTag0, s2, "<trash>", "<figure>\n\t\t\t\t<trash>",
output = writeField(buffer, s1, lastTag0, s2, "<trash>", "<trash>", addSpace, 3);
if (output) {
openFigure = true;
}
}
}
if (!output) {
output = writeField(buffer, s1, lastTag0, s2, "<equation>", "<formula>", addSpace, 3);
}
if (!output) {
output =
writeField(
buffer,
s1,
lastTag0,
s2,
"<figure_marker>",
"<ref type=\"figure\">",
addSpace,
3);
}
if (!output) {
if (openFigure) {
if (tableBlock && (!lastTag0.equals("<table>")) && (currentTag0.equals("<table>"))) {
buffer.append("\n\t\t\t</figure>\n\n");
output =
writeField(
buffer,
s1,
lastTag0,
s2,
"<figure>\n\t\t\t\t<table>",
"<figure>",
addSpace,
3);
if (output) {
tableBlock = true;
descFigure = false;
headFigure = false;
}
} else {
output = writeField(buffer, s1, lastTag0, s2, "<table>", "<table>", addSpace, 4);
if (output) {
tableBlock = true;
}
}
} else {
output =
writeField(
buffer, s1, lastTag0, s2, "<table>", "<figure>\n\t\t\t\t<table>", addSpace, 3);
if (output) {
openFigure = true;
tableBlock = true;
}
}
}
if (!output) {
if (openFigure) {
if (descFigure && (!lastTag0.equals("<label>")) && (currentTag0.equals("<label>"))) {
buffer.append("\n\t\t\t</figure>\n\n");
output =
writeField(
buffer,
s1,
lastTag0,
s2,
"<label>",
"<figure>\n\t\t\t\t<figDesc>",
addSpace,
3);
if (output) {
descFigure = true;
tableBlock = false;
headFigure = false;
}
} else {
output = writeField(buffer, s1, lastTag0, s2, "<label>", "<figDesc>", addSpace, 4);
if (output) {
descFigure = true;
}
}
} else {
output =
writeField(
buffer,
s1,
lastTag0,
s2,
"<label>",
"<figure>\n\t\t\t\t<figDesc>",
addSpace,
3);
if (output) {
openFigure = true;
descFigure = true;
}
}
}
if (!output) {
if (openFigure) {
if (headFigure
&& (!lastTag0.equals("<figure_head>"))
&& (currentTag0.equals("<figure_head>"))) {
buffer.append("\n\t\t\t</figure>\n\n");
output =
writeField(
buffer,
s1,
lastTag0,
s2,
"<figure_head>",
"<figure>\n\t\t\t\t<head>",
addSpace,
3);
if (output) {
descFigure = false;
tableBlock = false;
headFigure = true;
}
} else {
output = writeField(buffer, s1, lastTag0, s2, "<figure_head>", "<head>", addSpace, 4);
if (output) {
headFigure = true;
}
}
} else {
output =
writeField(
buffer,
s1,
lastTag0,
s2,
"<figure_head>",
"<figure>\n\t\t\t\t<head>",
addSpace,
3);
if (output) {
openFigure = true;
headFigure = true;
}
}
}
// for item we must distinguish starting and closing tags
if (!output) {
output = writeFieldBeginEnd(buffer, s1, lastTag, s2, "<item>", "<item>", addSpace, 3);
}
lastTag = s1;
if (!st.hasMoreTokens()) {
if (lastTag != null) {
testClosingTag(buffer, "", currentTag0, s1);
}
if (openFigure) {
buffer.append("\n\t\t\t</figure>\n\n");
}
}
if (start) {
start = false;
}
}
return buffer;
} catch (Exception e) {
throw new GrobidException("An exception occured while running Grobid.", e);
}
}
public static Document generalResultSegmentation(
Document doc, String labeledResult, List<LayoutToken> documentTokens) {
List<Pair<String, String>> labeledTokens = GenericTaggerUtils.getTokensAndLabels(labeledResult);
SortedSetMultimap<String, DocumentPiece> labeledBlocks = TreeMultimap.create();
doc.setLabeledBlocks(labeledBlocks);
/*try {
FileUtils.writeStringToFile(new File("/tmp/x1.txt"), labeledResult);
FileUtils.writeStringToFile(new File("/tmp/x2.txt"), documentTokens.toString());
}
catch(Exception e) {
e.printStackTrace();
}*/
List<Block> docBlocks = doc.getBlocks();
int indexLine = 0;
int blockIndex = 0;
int p = 0; // position in the labeled result
int currentLineEndPos = 0; // position in the global doc. tokenization of the last
// token of the current line
int currentLineStartPos = 0; // position in the global doc.
// tokenization of the first token of the current line
String line = null;
DocumentPointer pointerA = DocumentPointer.START_DOCUMENT_POINTER;
DocumentPointer currentPointer = null;
DocumentPointer lastPointer = null;
String curLabel;
String curPlainLabel = null;
String lastPlainLabel = null;
int lastTokenInd = -1;
for (int i = docBlocks.size() - 1; i >= 0; i--) {
int endToken = docBlocks.get(i).getEndToken();
if (endToken != -1) {
lastTokenInd = endToken;
break;
}
}
// we do this concatenation trick so that we don't have to process stuff after the main loop
// no copying of lists happens because of this, so it's ok to concatenate
String ignoredLabel = "@IGNORED_LABEL@";
for (Pair<String, String> labeledTokenPair :
Iterables.concat(
labeledTokens,
Collections.singleton(new Pair<String, String>("IgnoredToken", ignoredLabel)))) {
if (labeledTokenPair == null) {
p++;
continue;
}
// as we process the document segmentation line by line, we don't use the usual
// tokenization to rebuild the text flow, but we get each line again from the
// text stored in the document blocks (similarly as when generating the features)
line = null;
while ((line == null) && (blockIndex < docBlocks.size())) {
Block block = docBlocks.get(blockIndex);
List<LayoutToken> tokens = block.getTokens();
String localText = block.getText();
if ((tokens == null) || (localText == null) || (localText.trim().length() == 0)) {
blockIndex++;
indexLine = 0;
if (blockIndex < docBlocks.size()) {
block = docBlocks.get(blockIndex);
currentLineStartPos = block.getStartToken();
}
continue;
}
String[] lines = localText.split("[\\n\\r]");
if ((lines.length == 0) || (indexLine >= lines.length)) {
blockIndex++;
indexLine = 0;
if (blockIndex < docBlocks.size()) {
block = docBlocks.get(blockIndex);
currentLineStartPos = block.getStartToken();
}
continue;
} else {
line = lines[indexLine];
indexLine++;
if ((line.trim().length() == 0) || (TextUtilities.filterLine(line))) {
line = null;
continue;
}
if (currentLineStartPos > lastTokenInd) continue;
// adjust the start token position in documentTokens to this non trivial line
// first skip possible space characters and tabs at the beginning of the line
while ((documentTokens.get(currentLineStartPos).t().equals(" ")
|| documentTokens.get(currentLineStartPos).t().equals("\t"))
&& (currentLineStartPos != lastTokenInd)) {
currentLineStartPos++;
}
if (!labeledTokenPair.a.startsWith(documentTokens.get(currentLineStartPos).getText())) {
while (currentLineStartPos < block.getEndToken()) {
if (documentTokens.get(currentLineStartPos).t().equals("\n")
|| documentTokens.get(currentLineStartPos).t().equals("\r")) {
// move to the start of the next line, but ignore space characters and tabs
currentLineStartPos++;
while ((documentTokens.get(currentLineStartPos).t().equals(" ")
|| documentTokens.get(currentLineStartPos).t().equals("\t"))
&& (currentLineStartPos != lastTokenInd)) {
currentLineStartPos++;
}
if ((currentLineStartPos != lastTokenInd)
&& labeledTokenPair.a.startsWith(
documentTokens.get(currentLineStartPos).getText())) {
break;
}
}
currentLineStartPos++;
}
}
// what is then the position of the last token of this line?
currentLineEndPos = currentLineStartPos;
while (currentLineEndPos < block.getEndToken()) {
if (documentTokens.get(currentLineEndPos).t().equals("\n")
|| documentTokens.get(currentLineEndPos).t().equals("\r")) {
currentLineEndPos--;
break;
}
currentLineEndPos++;
}
}
}
curLabel = labeledTokenPair.b;
curPlainLabel = GenericTaggerUtils.getPlainLabel(curLabel);
/*System.out.println("-------------------------------");
System.out.println("block: " + blockIndex);
System.out.println("line: " + line);
System.out.println("token: " + labeledTokenPair.a);
System.out.println("curPlainLabel: " + curPlainLabel);
System.out.println("lastPlainLabel: " + lastPlainLabel);
if ((currentLineStartPos < lastTokenInd) && (currentLineStartPos != -1))
System.out.println("currentLineStartPos: " + currentLineStartPos +
" (" + documentTokens.get(currentLineStartPos) + ")");
if ((currentLineEndPos < lastTokenInd) && (currentLineEndPos != -1))
System.out.println("currentLineEndPos: " + currentLineEndPos +
" (" + documentTokens.get(currentLineEndPos) + ")");*/
if (blockIndex == docBlocks.size()) {
break;
}
currentPointer = new DocumentPointer(doc, blockIndex, currentLineEndPos);
// either a new entity starts or a new beginning of the same type of entity
if ((!curPlainLabel.equals(lastPlainLabel)) && (lastPlainLabel != null)) {
if ((pointerA.getTokenDocPos() <= lastPointer.getTokenDocPos())
&& (pointerA.getTokenDocPos() != -1)) {
labeledBlocks.put(lastPlainLabel, new DocumentPiece(pointerA, lastPointer));
}
pointerA = new DocumentPointer(doc, blockIndex, currentLineStartPos);
// System.out.println("add segment for: " + lastPlainLabel + ", until " +
// (currentLineStartPos-2));
}
// updating stuff for next iteration
lastPlainLabel = curPlainLabel;
lastPointer = currentPointer;
currentLineStartPos = currentLineEndPos + 2; // one shift for the EOL, one for the next line
p++;
}
if (blockIndex == docBlocks.size()) {
// the last labelled piece has still to be added
if ((!curPlainLabel.equals(lastPlainLabel)) && (lastPlainLabel != null)) {
if ((pointerA.getTokenDocPos() <= lastPointer.getTokenDocPos())
&& (pointerA.getTokenDocPos() != -1)) {
labeledBlocks.put(lastPlainLabel, new DocumentPiece(pointerA, lastPointer));
// System.out.println("add segment for: " + lastPlainLabel + ", until " +
// (currentLineStartPos-2));
}
}
}
return doc;
}
/**
* First pass to detect basic structures: remove page header/footer, identify section numbering,
* identify Figure and table blocks.
*
* <p>-> to be removed at some point!
*
* @param doc a document
*/
public static void firstPass(Document doc) {
if (doc == null) {
throw new NullPointerException();
}
if (doc.getBlocks() == null) {
throw new NullPointerException();
}
int i = 0;
List<Integer> blockHeaders = new ArrayList<Integer>();
List<Integer> blockFooters = new ArrayList<Integer>();
List<Integer> blockSectionTitles = new ArrayList<Integer>();
List<Integer> acknowledgementBlocks = new ArrayList<Integer>();
List<Integer> blockTables = new ArrayList<Integer>();
List<Integer> blockFigures = new ArrayList<Integer>();
List<Integer> blockHeadTables = new ArrayList<Integer>();
List<Integer> blockHeadFigures = new ArrayList<Integer>();
List<Integer> blockDocumentHeaders = new ArrayList<Integer>();
doc.setTitleMatchNum(false);
try {
for (Block block : doc.getBlocks()) {
String localText = block.getText().trim();
localText = localText.replace("\n", " ");
localText = localText.replace(" ", " ");
localText = localText.trim();
Matcher ma1 = BasicStructureBuilder.introduction.matcher(localText);
Matcher ma2 = BasicStructureBuilder.references.matcher(localText);
if ((ma1.find()) || (ma2.find())) {
if (((localText.startsWith("1.")) || (localText.startsWith("1 ")))
|| ((localText.startsWith("2.")) || (localText.startsWith("2 ")))
|| (localText.startsWith("Contents"))) doc.setTitleMatchNum(true);
// System.out.println("Title section identified: block " + i + ", " + localText);
blockSectionTitles.add(i);
} else {
StringTokenizer st = new StringTokenizer(localText, "\n");
while (st.hasMoreTokens()) {
String token = st.nextToken();
if (token.startsWith("@PAGE")) {
// current block should give the header/footors
if (i > 4) {
if (doc.getBlocks().get(i - 5).getNbTokens() < 20) {
Integer i2 = i - 5;
if (!blockFooters.contains(i2)) blockFooters.add(i2);
}
}
if (i > 3) {
if (doc.getBlocks().get(i - 4).getNbTokens() < 20) {
Integer i2 = i - 4;
if (!blockFooters.contains(i2)) blockFooters.add(i2);
}
}
if (i > 2) {
if (doc.getBlocks().get(i - 3).getNbTokens() < 20) {
Integer i2 = i - 3;
if (!blockFooters.contains(i2)) blockFooters.add(i2);
}
}
if (i > 1) {
if (doc.getBlocks().get(i - 2).getNbTokens() < 20) {
Integer i2 = i - 2;
if (!blockFooters.contains(i2)) blockFooters.add(i2);
}
}
if (i > 0) {
if (doc.getBlocks().get(i - 1).getNbTokens() < 20) {
Integer i2 = i - 1;
if (!blockFooters.contains(i2)) blockFooters.add(i2);
}
}
blockFooters.add(i);
// page header candidates
blockHeaders.add(i);
if (i < doc.getBlocks().size() - 1) {
if (doc.getBlocks().get(i + 1).getNbTokens() < 20) {
Integer i2 = i + 1;
if (!blockHeaders.contains(i2)) blockHeaders.add(i + 1);
}
}
if (i < doc.getBlocks().size() - 2) {
if (doc.getBlocks().get(i + 2).getNbTokens() < 20) {
Integer i2 = i + 2;
if (!blockHeaders.contains(i2)) blockHeaders.add(i + 2);
}
}
if (i < doc.getBlocks().size() - 3) {
if (doc.getBlocks().get(i + 3).getNbTokens() < 20) {
Integer i2 = i + 3;
if (!blockHeaders.contains(i2)) blockHeaders.add(i + 3);
}
}
if (i < doc.getBlocks().size() - 4) {
if (doc.getBlocks().get(i + 4).getNbTokens() < 20) {
Integer i2 = i + 4;
if (!blockHeaders.contains(i2)) blockHeaders.add(i + 4);
}
}
// more ??
}
}
}
// clustering of blocks per font (for section header and figure/table detections)
addBlockToCluster(i, doc);
i++;
}
// try to find the cluster of section titles
Cluster candidateCluster = null;
// System.out.println("nb clusters: " + clusters.size());
for (Cluster cluster : doc.getClusters()) {
if ((cluster.getNbBlocks() < (doc.getBlocks().size() / 5))
&& (cluster.getNbBlocks() < 20)) {
List<Integer> blo = cluster.getBlocks2();
for (Integer b : blo) {
if (blockSectionTitles.contains(b)) {
if (candidateCluster == null) {
candidateCluster = cluster;
break;
}
// else if (cluster.getFontSize() >= candidateCluster.getFontSize())
// candidateCluster = cluster;
}
}
}
}
if (candidateCluster != null) {
List<Integer> newBlockSectionTitles = new ArrayList<Integer>();
for (Integer bl : blockSectionTitles) {
if (!newBlockSectionTitles.contains(bl)) newBlockSectionTitles.add(bl);
}
List<Integer> blockClusterTitles = candidateCluster.getBlocks2();
if (blockClusterTitles.size() < 20) {
for (Integer bl : blockClusterTitles) {
if (!newBlockSectionTitles.contains(bl)) newBlockSectionTitles.add(bl);
}
}
blockSectionTitles = newBlockSectionTitles;
}
// aknowledgement section recognition
boolean ackn = false;
i = 0;
for (Block block : doc.getBlocks()) {
String localText = block.getText().trim();
localText = localText.replace("\n", " ");
localText = localText.replace(" ", " ");
localText = localText.trim();
// System.out.println(i + ": " + localText+"\n");
Integer iii = i;
Matcher m3 = BasicStructureBuilder.acknowledgement.matcher(localText);
if ((m3.find()) && (blockSectionTitles.contains(iii))) {
acknowledgementBlocks.add(iii);
ackn = true;
// int index = blockSectionTitles.indexOf(iii);
// blockSectionTitles.remove(index);
} else if ((ackn) && (blockSectionTitles.contains(iii))) {
ackn = false;
break;
} else if (ackn) {
Matcher m4 = BasicStructureBuilder.references.matcher(localText);
if ((ackn) && (!blockFooters.contains(iii)) && (!m4.find())) {
acknowledgementBlocks.add(iii);
} else if (m4.find()) {
ackn = false;
break;
}
}
i++;
}
// we remove references headers in blockSectionTitles
int index = -1;
for (Integer ii : blockSectionTitles) {
Block block = doc.getBlocks().get(ii);
String localText = block.getText().trim();
localText = localText.replace("\n", " ");
localText = localText.replace(" ", " ");
localText = localText.trim();
Matcher m4 = BasicStructureBuilder.references.matcher(localText);
if (m4.find()) {
index = blockSectionTitles.indexOf(ii);
break;
}
}
if (index != -1) {
blockSectionTitles.remove(index);
}
// we check headers repetition from page to page to decide if it is an header or not
ArrayList<Integer> toRemove = new ArrayList<Integer>();
for (Integer ii : blockHeaders) {
String localText = (doc.getBlocks().get(ii)).getText().trim();
localText = TextUtilities.shadowNumbers(localText);
int length = localText.length();
if (length > 160) toRemove.add(ii);
else {
// System.out.println("header candidate: " + localText);
// evaluate distance with other potential headers
boolean valid = false;
for (Integer ii2 : blockHeaders) {
if (ii.intValue() != ii2.intValue()) {
String localText2 = doc.getBlocks().get(ii2).getText().trim();
if (localText2.length() < 160) {
localText2 = TextUtilities.shadowNumbers(localText2);
double dist =
(double) TextUtilities.getLevenshteinDistance(localText, localText2) / length;
// System.out.println("dist with " + localText2 + " : " + dist);
if (dist < 0.25) {
valid = true;
break;
}
}
}
}
if (!valid) {
toRemove.add(ii);
}
}
}
for (Integer ii : toRemove) {
blockHeaders.remove(ii);
}
// same for footers
toRemove = new ArrayList<Integer>();
for (Integer ii : blockFooters) {
String localText = (doc.getBlocks().get(ii)).getText().trim();
localText = TextUtilities.shadowNumbers(localText);
int length = localText.length();
if (length > 160) toRemove.add(ii);
else {
// System.out.println("footer candidate: " + localText);
// evaluate distance with other potential headers
boolean valid = false;
for (Integer ii2 : blockFooters) {
if (ii.intValue() != ii2.intValue()) {
String localText2 = doc.getBlocks().get(ii2).getText().trim();
if (localText2.length() < 160) {
localText2 = TextUtilities.shadowNumbers(localText2);
double dist =
(double) TextUtilities.getLevenshteinDistance(localText, localText2) / length;
if (dist < 0.25) {
valid = true;
break;
}
}
}
}
if (!valid) {
toRemove.add(ii);
}
}
}
for (Integer ii : toRemove) {
blockFooters.remove(ii);
}
// a special step for added banner repositoryies such HAL
i = 0;
for (Block block : doc.getBlocks()) {
String localText = block.getText().trim();
localText = localText.replace("\n", " ");
localText = localText.replace(" ", " ");
localText = localText.trim();
// HAL
if (localText.startsWith("Author manuscript, published in")) {
Double y = block.getY();
// System.out.println("HAL banner found, " + "block " + i + ", y = " + y);
if (Math.abs(y - 12.538) < 2) { // reference position
// blockHeaders.add(new Integer(i));
blockDocumentHeaders.add(i);
// System.out.println("HAL banner added as header block");
break;
}
}
// ACM publications
// System.out.println("test ACM " + i);
// System.out.println(localText);
if (localText.startsWith("Permission to make digital or hard copies")) {
blockFooters.add(i);
break;
}
// arXiv, etc. put here
// IOP
if (localText.startsWith("Confidential: ") && (localText.contains("IOP"))) {
blockDocumentHeaders.add(i);
// System.out.println("IOP banner added as header block");
break;
}
i++;
}
// we try to recognize here table and figure blocks
// the idea is that the textual elements are not located as the normal text blocks
// this is recognized by exploiting the cluster of blocks starting up and down front the block
// containing a table or a figure marker
// two different runs, one for figures and one for tables (everything could be done in one
// step)
i = 0;
for (Block block : doc.getBlocks()) {
String localText = block.getText().trim();
localText = localText.replace("\n", " ");
localText = localText.replace(" ", " ");
localText = localText.trim();
Matcher m = BasicStructureBuilder.figure.matcher(localText);
Matcher m2 = BasicStructureBuilder.table.matcher(localText);
double width = block.getWidth();
boolean bold = block.getBold();
// table
// if ( (m2.find()) && (localText.length() < 200) ) {
if ((m2.find()) && ((bold) || (localText.length() < 200))) {
if (!blockHeadTables.contains(i)) {
blockHeadTables.add(i);
}
// we also put all the small blocks before and after the marker
int j = i - 1;
while ((j > i - 15) && (j > 0)) {
Block b = doc.getBlocks().get(j);
if (b.getText() != null) {
if ((b.getText().length() < 160) || (width < 50)) {
if ((!blockTables.contains(j))
&& (!blockSectionTitles.contains(j))
&& (!blockHeaders.contains(j))
&& (!blockFooters.contains(j))) blockTables.add(j);
} else j = 0;
}
j--;
}
j = i + 1;
while ((j < i + 15) && (j < doc.getBlocks().size())) {
Block b = doc.getBlocks().get(j);
if (b.getText() != null) {
if ((b.getText().length() < 160) || (width < 50)) {
if ((!blockTables.contains(j))
&& (!blockSectionTitles.contains(j))
&& (!blockHeaders.contains(j))
&& (!blockFooters.contains(j))) blockTables.add(j);
} else j = doc.getBlocks().size();
}
j++;
}
}
// figure
// else if ( (m.find()) && (localText.length() < 200) ) {
else if ((m.find()) && ((bold) || (localText.length() < 200))) {
if (!blockHeadFigures.contains(i)) blockHeadFigures.add(i);
// we also put all the small blocks before and after the marker
int j = i - 1;
boolean imageFound = false;
while ((j > i - 15) && (j > 0)) {
Block b = doc.getBlocks().get(j);
if (b.getText() != null) {
String localText2 = b.getText().trim();
// localText = localText.replace("\n", " ");
localText2 = localText2.replace(" ", " ");
localText2 = localText2.trim();
if ((localText2.startsWith("@IMAGE")) && (!imageFound)) {
// System.out.println(localText2);
block.setText(block.getText() + " " + localText2);
// System.out.println(block.getText());
imageFound = true;
}
if ((localText2.length() < 160) || (width < 50)) {
if ((!blockFigures.contains(j))
&& (!blockSectionTitles.contains(j))
&& (!blockHeaders.contains(j))
&& (!blockFooters.contains(j))) blockFigures.add(j);
} else j = 0;
}
j--;
}
j = i + 1;
while ((j < i + 15) && (j < doc.getBlocks().size())) {
Block b = doc.getBlocks().get(j);
if (b.getText() != null) {
if ((b.getText().trim().length() < 160) || (width < 50)) {
if ((!blockFigures.contains(j))
&& (!blockSectionTitles.contains(j))
&& (!blockHeaders.contains(j))
&& (!blockFooters.contains(j))) blockFigures.add(j);
} else j = doc.getBlocks().size();
}
j++;
}
}
i++;
}
} finally {
doc.setBlockHeaders(blockHeaders);
doc.setBlockFooters(blockFooters);
doc.setBlockSectionTitles(blockSectionTitles);
doc.setAcknowledgementBlocks(acknowledgementBlocks);
doc.setBlockTables(blockTables);
doc.setBlockFigures(blockFigures);
doc.setBlockHeadTables(blockHeadTables);
doc.setBlockHeadFigures(blockHeadFigures);
doc.setBlockDocumentHeaders(blockDocumentHeaders);
}
}
public String printVector() {
if (string == null) return null;
if (string.length() == 0) return null;
StringBuilder res = new StringBuilder();
// token string (1)
res.append(string);
// lowercase string (1)
res.append(" ").append(string.toLowerCase());
// prefix (4)
res.append(" " + TextUtilities.prefix(string, 1));
res.append(" " + TextUtilities.prefix(string, 2));
res.append(" " + TextUtilities.prefix(string, 3));
res.append(" " + TextUtilities.prefix(string, 4));
// suffix (4)
res.append(" " + TextUtilities.suffix(string, 1));
res.append(" " + TextUtilities.suffix(string, 2));
res.append(" " + TextUtilities.suffix(string, 3));
res.append(" " + TextUtilities.suffix(string, 4));
// line information (1)
res.append(" ").append(lineStatus);
// capitalisation (1)
if (digit.equals("ALLDIGIT")) res.append(" NOCAPS");
else res.append(" ").append(capitalisation);
// digit information (1)
res.append(" ").append(digit);
// character information (1)
if (singleChar) res.append(" 1");
else res.append(" 0");
// lexical information (8)
if (properName) res.append(" 1");
else res.append(" 0");
if (commonName) res.append(" 1");
else res.append(" 0");
if (firstName) res.append(" 1");
else res.append(" 0");
if (locationName) res.append(" 1");
else res.append(" 0");
if (year) res.append(" 1");
else res.append(" 0");
if (month) res.append(" 1");
else res.append(" 0");
if (email) res.append(" 1");
else res.append(" 0");
if (http) res.append(" 1");
else res.append(" 0");
// bibliographical information(4)
if (isKnownJournalTitle || isKnownAbbrevJournalTitle) res.append(" 1");
else res.append(" 0");
if (isKnownConferenceTitle) res.append(" 1");
else res.append(" 0");
if (isKnownPublisher) res.append(" 1");
else res.append(" 0");
// punctuation information (1)
res.append(" ").append(punctType); // in case the token is a punctuation (NO otherwise)
// relative position in the sequence (1)
res.append(" ").append(relativePosition);
// label - for training data (1)
if (label != null) res.append(" ").append(label).append("\n");
else res.append(" 0\n");
return res.toString();
}
public void endElement(java.lang.String uri, java.lang.String localName, java.lang.String qName)
throws SAXException {
// if (!qName.equals("TOKEN") && !qName.equals("BLOCK") &&
// !qName.equals("TEXT"))
// System.out.println(qName);
if (qName.equals("TEXT")) {
blabla.append("\n");
LayoutToken token = new LayoutToken();
token.setText("\n");
block.addToken(token);
accumulator.setLength(0);
tokenizations.add("\n");
} else if (qName.equals("METADATA")) {
accumulator.setLength(0);
} else if (qName.equals("TOKEN")) {
String tok0 = TextUtilities.clean(getText());
if (block.getStartToken() == -1) {
block.setStartToken(tokenizations.size());
}
if (tok0.length() > 0) {
StringTokenizer st = new StringTokenizer(tok0, TextUtilities.fullPunctuations, true);
boolean diaresis = false;
boolean accent = false;
boolean keepLast = false;
while (st.hasMoreTokens()) {
diaresis = false;
accent = false;
keepLast = false;
String tok = st.nextToken();
if (tok.length() > 0) {
LayoutToken token = new LayoutToken();
if ((previousToken != null)
&& (tok != null)
&& (previousToken.length() > 0)
&& (tok.length() > 0)
&& blabla.length() > 0) {
Character leftChar = previousTok.getText().charAt(previousTok.getText().length() - 1);
Character rightChar = tok.charAt(0);
ModifierClass leftClass = classifyChar(leftChar);
ModifierClass rightClass = classifyChar(rightChar);
ModifierClass modifierClass = ModifierClass.NOT_A_MODIFIER;
if (leftClass != ModifierClass.NOT_A_MODIFIER
|| rightClass != ModifierClass.NOT_A_MODIFIER) {
Character baseChar = null;
Character modifierChar = null;
if (leftClass != ModifierClass.NOT_A_MODIFIER) {
if (rightClass != ModifierClass.NOT_A_MODIFIER) {
// assert false;
// keeping characters, but setting class
// to not a modifier
baseChar = leftChar;
modifierChar = rightChar;
modifierClass = ModifierClass.NOT_A_MODIFIER;
} else {
baseChar = rightChar;
modifierChar = leftChar;
modifierClass = leftClass;
}
} else {
baseChar = leftChar;
modifierChar = rightChar;
modifierClass = rightClass;
}
String updatedChar = modifyCharacter(baseChar, modifierChar);
tokenizations.remove(tokenizations.size() - 1);
if (tokenizations.size() > 0) {
tokenizations.remove(tokenizations.size() - 1);
}
blabla.deleteCharAt(blabla.length() - 1);
if (blabla.length() > 0) {
blabla.deleteCharAt(blabla.length() - 1);
}
removeLastCharacterIfPresent(previousTok);
if (updatedChar != null) {
blabla.append(updatedChar);
previousTok.setText(previousTok.getText() + updatedChar);
}
blabla.append(tok.substring(1, tok.length()));
previousTok.setText(previousTok.getText() + tok.substring(1, tok.length()));
tokenizations.add(previousTok.getText());
diaresis =
(modifierClass == ModifierClass.DIAERESIS
|| modifierClass == ModifierClass.NORDIC_RING
|| modifierClass == ModifierClass.CZECH_CARON
|| modifierClass == ModifierClass.TILDE
|| modifierClass == ModifierClass.CEDILLA);
accent =
(modifierClass == ModifierClass.ACUTE_ACCENT
|| modifierClass == ModifierClass.CIRCUMFLEX
|| modifierClass == ModifierClass.GRAVE_ACCENT);
if (rightClass != ModifierClass.NOT_A_MODIFIER) {
tok = ""; // resetting current token as it
// is a single-item
}
}
}
if (tok != null) {
// actually in certain cases, the extracted string under token can be a chunk of text
// with separators that need to be preserved
// tok = tok.replace(" ", "");
}
if ((!diaresis) && (!accent)) {
// blabla.append(" ");
blabla.append(tok);
token.setText(tok);
tokenizations.add(tok);
} else {
tok = "";
keepLast = true;
}
/*
* StringTokenizer st0 = new StringTokenizer(tok0,
* TextUtilities.fullPunctuations, true);
* while(st0.hasMoreTokens()) { String tok =
* st0.nextToken(); tokenizations.add(tok); }
* tokenizations.add(" ");
*/
/*
* boolean punct1 = false; boolean punct2 = false;
* boolean punct3 = false; String content = null; int i
* = 0; for(; i<TextUtilities.punctuations.length();
* i++) { if (tok.length() > 0) { if
* (tok.charAt(tok.length()-1) ==
* TextUtilities.punctuations.charAt(i)) { punct1 =
* true; content = tok.substring(0, tok.length()-1); if
* (tok.length() > 1) { int j = 0; for(;
* j<TextUtilities.punctuations.length(); j++) { if
* (tok.charAt(tok.length()-2) ==
* TextUtilities.punctuations.charAt(j)) { punct3 =
* true; content = tok.substring(0, tok.length()-2); } }
* } break; } } } if (tok.length() > 0) { if (
* (tok.startsWith("(")) && (tok.length() > 1) ) { if
* ((punct3) && (tok.length() > 2)) content =
* tok.substring(1, tok.length()-2); else if (punct1)
* content = tok.substring(1, tok.length()-1); else
* content = tok.substring(1, tok.length()); punct2 =
* true; token.setText("("); } else if (
* (tok.startsWith("[")) && (tok.length() > 1) ) { if
* ((punct3) && (tok.length() > 2)) content =
* tok.substring(1, tok.length()-2); else if (punct1)
* content = tok.substring(1, tok.length()-1); else
* content = tok.substring(1, tok.length()); punct2 =
* true; token.setText("["); } else if (
* (tok.startsWith("\"")) && (tok.length() > 1) ) { if
* ((punct3) && (tok.length() > 2)) content =
* tok.substring(1, tok.length()-2); else if (punct1)
* content = tok.substring(1, tok.length()-1); else
* content = tok.substring(1, tok.length()); punct2 =
* true; token.setText("\""); } }
*/
if (currentRotation) currentFontSize = currentFontSize / 2;
/*
* if (punct2) { if (currentFont != null)
* token.setFont(currentFont.toLowerCase()); else
* token.setFont("default");
* token.setItalic(currentItalic);
* token.setBold(currentBold);
* token.setRotation(currentRotation);
* token.setColorFont(colorFont); token.setX(currentX);
* token.setY(currentY); token.setWidth(currentWidth);
* token.setHeight(currentHeight);
* token.setFontSize(currentFontSize);
* block.addToken(token);
*
* token = new LayoutToken(); token.setText(content); }
* if (punct1) { token.setText(content); if (currentFont
* != null) token.setFont(currentFont.toLowerCase());
* else token.setFont("default");
* token.setItalic(currentItalic);
* token.setBold(currentBold);
* token.setRotation(currentRotation);
* token.setColorFont(colorFont); token.setX(currentX);
* token.setY(currentY); token.setWidth(currentWidth);
* token.setHeight(currentHeight);
* token.setFontSize(currentFontSize);
* block.addToken(token);
*
* if (punct3) { token = new LayoutToken();
* token.setText(""+tok.charAt(tok.length()-2)); if
* (currentFont != null)
* token.setFont(currentFont.toLowerCase()); else
* token.setFont("default");
* token.setItalic(currentItalic);
* token.setBold(currentBold);
* token.setRotation(currentRotation);
* token.setColorFont(colorFont); token.setX(currentX);
* token.setY(currentY); token.setWidth(currentWidth);
* token.setHeight(currentHeight);
* token.setFontSize(currentFontSize);
* block.addToken(token); }
*
* token = new LayoutToken();
* token.setText(""+tok.charAt(tok.length()-1)); }
*/
if (currentFont != null) token.setFont(currentFont.toLowerCase());
else token.setFont("default");
token.setItalic(currentItalic);
token.setBold(currentBold);
token.setRotation(currentRotation);
token.setColorFont(colorFont);
token.setX(currentX);
token.setY(currentY);
token.setWidth(currentWidth);
token.setHeight(currentHeight);
token.setFontSize(currentFontSize);
if (!diaresis && !accent) {
block.addToken(token);
}
if (block.getFont() == null) {
if (currentFont != null) block.setFont(currentFont.toLowerCase());
else token.setFont("default");
}
if (nbTokens == 0) {
block.setItalic(currentItalic);
block.setBold(currentBold);
}
if (block.getColorFont() == null) block.setColorFont(colorFont);
if (block.getX() == 0.0) block.setX(currentX);
if (block.getY() == 0.0) block.setY(currentY);
if (block.getWidth() == 0.0) block.setWidth(currentWidth);
if (block.getHeight() == 0.0) block.setHeight(currentHeight);
if (block.getFontSize() == 0.0) block.setFontSize(currentFontSize);
if (!diaresis && !accent) {
previousToken = tok;
previousTok = token;
} else {
previousToken = previousTok.getText();
}
nbTokens++;
accumulator.setLength(0);
}
}
if (tokenizations.size() > 0) {
String justBefore = tokenizations.get(tokenizations.size() - 1);
if (!justBefore.endsWith("-")) {
tokenizations.add(" ");
blabla.append(" ");
}
}
}
block.setEndToken(tokenizations.size());
} else if (qName.equals("PAGE")) {
// page marker are usefull to detect headers (same first line(s)
// appearing on each page)
if (block != null) {
blabla.append("\n");
tokenizations.add("\n");
block.setText(blabla.toString());
block.setNbTokens(nbTokens);
doc.addBlock(block);
}
Block block0 = new Block();
block0.setText("@PAGE\n");
block0.setNbTokens(0);
block0.setPage(currentPage);
doc.addBlock(block0);
block = new Block();
block.setPage(currentPage);
blabla = new StringBuffer();
nbTokens = 0;
// blabla.append("\n@block\n");
tokenizations.add("\n");
} else if (qName.equals("IMAGE")) {
if (block != null) {
blabla.append("\n");
block.setText(blabla.toString());
block.setNbTokens(nbTokens);
doc.addBlock(block);
}
block = new Block();
block.setPage(currentPage);
blabla = new StringBuffer();
if (images.size() > 0) {
blabla.append("@IMAGE " + images.get(images.size() - 1) + "\n");
}
block.setText(blabla.toString());
block.setNbTokens(nbTokens);
if (block.getX() == 0.0) block.setX(currentX);
if (block.getY() == 0.0) block.setY(currentY);
if (block.getWidth() == 0.0) block.setWidth(currentWidth);
if (block.getHeight() == 0.0) block.setHeight(currentHeight);
doc.addBlock(block);
blabla = new StringBuffer();
nbTokens = 0;
block = new Block();
block.setPage(currentPage);
}
/*
* else if (qName.equals("VECTORIALIMAGES")) { if (block != null) {
* blabla.append("\n"); block.setText(blabla.toString());
* block.setNbTokens(nbTokens); doc.addBlock(block); } block = new
* Block(); block.setPage(currentPage); blabla = new StringBuffer();
* blabla.append("@IMAGE " + "vectorial \n");
* block.setText(blabla.toString()); block.setNbTokens(nbTokens); if
* (block.getX() == 0.0) block.setX(currentX); if (block.getY() == 0.0)
* block.setY(currentY); if (block.getWidth() == 0.0)
* block.setWidth(currentWidth); if (block.getHeight() == 0.0)
* block.setHeight(currentHeight); doc.addBlock(block); blabla = new
* StringBuffer(); nbTokens = 0; block = new Block();
* block.setPage(currentPage); }
*/
else if (qName.equals("BLOCK")) {
blabla.append("\n");
tokenizations.add("\n");
block.setText(blabla.toString());
block.setNbTokens(nbTokens);
block.setWidth(currentX - block.getX() + currentWidth);
block.setHeight(currentY - block.getY() + currentHeight);
doc.addBlock(block);
// blabla = new StringBuffer();
nbTokens = 0;
block = null;
} else if (qName.equals("xi:include")) {
if (block != null) {
blabla.append("\n");
block.setText(blabla.toString());
block.setNbTokens(nbTokens);
doc.addBlock(block);
}
block = new Block();
block.setPage(currentPage);
blabla = new StringBuffer();
blabla.append("@IMAGE " + images.get(images.size() - 1) + "\n");
block.setText(blabla.toString());
block.setNbTokens(nbTokens);
doc.addBlock(block);
blabla = new StringBuffer();
nbTokens = 0;
block = new Block();
block.setPage(currentPage);
}
/*
* else if (qName.equals("DOCUMENT")) {
* System.out.println(blabla.toString()); }
*/
}