/**
* tokenizes a given JCas object's document text using the treetagger program and adds the
* recognized tokens to the JCas object.
*
* @param jcas JCas object supplied by the pipeline
*/
private void tokenize(JCas jcas) {
// read tokenized text to add tokens to the jcas
Logger.printDetail(component, "TreeTagger (tokenization) with: " + ttprops.abbFileName);
EnumSet<Flag> flags = Flag.getSet(ttprops.languageSwitch);
TreeTaggerTokenizer ttt;
ttprops.abbFileName = "english-abbreviations";
if (ttprops.abbFileName != null) {
ttt =
new TreeTaggerTokenizer(
ttprops.rootPath
+ ttprops.fileSeparator
+ "lib"
+ ttprops.fileSeparator
+ ttprops.abbFileName,
flags);
} else {
ttt = new TreeTaggerTokenizer(null, flags);
}
String docText = jcas.getDocumentText().replaceAll("\n\n", "\nEMPTYLINE\n");
List<String> tokenized = ttt.tokenize(docText);
int tokenOffset = 0;
// loop through all the lines in the treetagger output
for (String s : tokenized) {
// charset missmatch fallback: signal (invalid) s
if ((!(s.equals("EMPTYLINE"))) && (jcas.getDocumentText().indexOf(s, tokenOffset) < 0))
throw new RuntimeException(
"Opps! Could not find token "
+ s
+ " in JCas after tokenizing with TreeTagger."
+ " Hmm, there may exist a charset missmatch!"
+ " Default encoding is "
+ Charset.defaultCharset().name()
+ " and should always be UTF-8 (use -Dfile.encoding=UTF-8)."
+ " If input document is not UTF-8 use -e option to set it according to the input, additionally.");
// create tokens and add them to the jcas's indexes.
Token newToken = new Token(jcas);
if (s.equals("EMPTYLINE")) {
newToken.setBegin(tokenOffset);
newToken.setEnd(tokenOffset);
newToken.setPos("EMPTYLINE");
if (annotate_partofspeech) {
newToken.addToIndexes();
}
} else {
newToken.setBegin(jcas.getDocumentText().indexOf(s, tokenOffset));
newToken.setEnd(newToken.getBegin() + s.length());
newToken.addToIndexes();
tokenOffset = newToken.getEnd();
}
}
}
/**
* tokenizes a given JCas object's document text using the chinese tokenization script and adds
* the recognized tokens to the JCas object.
*
* @param jcas JCas object supplied by the pipeline
*/
private void tokenizeChinese(JCas jcas) {
try {
// read tokenized text to add tokens to the jcas
Process proc = ttprops.getChineseTokenizationProcess();
Logger.printDetail(component, "Chinese tokenization: " + ttprops.chineseTokenizerPath);
BufferedReader in = new BufferedReader(new InputStreamReader(proc.getInputStream(), "UTF-8"));
BufferedWriter out =
new BufferedWriter(new OutputStreamWriter(proc.getOutputStream(), "UTF-8"));
Integer tokenOffset = 0;
// loop through all the lines in the stdout output
String[] inSplits = jcas.getDocumentText().split("[\\r\\n]+");
for (String inSplit : inSplits) {
out.write(inSplit);
out.newLine();
out.flush();
// do one initial read
String s = in.readLine();
do {
// break out of the loop if we've read a null
if (s == null) break;
String[] outSplits = s.split("\\s+");
for (String tok : outSplits) {
if (jcas.getDocumentText().indexOf(tok, tokenOffset) < 0)
throw new RuntimeException(
"Could not find token "
+ tok
+ " in JCas after tokenizing with Chinese tokenization script.");
// create tokens and add them to the jcas's indexes.
Token newToken = new Token(jcas);
newToken.setBegin(jcas.getDocumentText().indexOf(tok, tokenOffset));
newToken.setEnd(newToken.getBegin() + tok.length());
newToken.addToIndexes();
tokenOffset = newToken.getEnd();
}
// break out of the loop if the next read will block
if (!in.ready()) break;
s = in.readLine();
} while (true);
}
// clean up
in.close();
proc.destroy();
} catch (Exception e) {
e.printStackTrace();
}
}
