@Override
public Vector<Double> calculateScores(JCas aCas) throws ScoringComponentException {
// 1) how many words of H (extended with multiple relations) can be
// found in T divided by the length of H
Vector<Double> scoresVector = new Vector<Double>();
try {
JCas tView = aCas.getView("TextView");
HashMap<String, Integer> tBag = countTokens(tView);
JCas hView = aCas.getView("HypothesisView");
HashMap<String, Integer> hBag = countTokens(hView);
if (null != wnlrSet && wnlrSet.size() != 0) {
for (WordnetLexicalResource wnlr : wnlrSet) {
scoresVector.add(calculateSingleLexScoreWithWNRelations(tBag, hBag, wnlr));
}
}
if (null != volrSet && volrSet.size() != 0) {
for (VerbOceanLexicalResource volr : volrSet) {
scoresVector.add(calculateSingleLexScoreWithVORelations(tBag, hBag, volr));
}
}
} catch (CASException e) {
throw new ScoringComponentException(e.getMessage());
}
return scoresVector;
}
@Override
public Vector<Double> calculateScores(JCas aCas) throws ScoringComponentException {
// all the values: (T&H/H), (T&H/T), and ((T&H/H)*(T&H/T)), with four
// different matching types
Vector<Double> scoresVector = new Vector<Double>();
try {
JCas tView = aCas.getView("TextView");
JCas hView = aCas.getView("HypothesisView");
for (int i = 1; i < 5; i++) {
HashMap<String, Integer> tBag = countDeps(tView, i);
HashMap<String, Integer> hBag = countDeps(hView, i);
scoresVector.addAll(calculateSimilarity(tBag, hBag));
}
} catch (CASException e) {
throw new ScoringComponentException(e.getMessage());
}
return scoresVector;
}
public void entityProcessComplete(CAS aCas, EntityProcessStatus aStatus) {
if (aStatus != null) {
if (aStatus.isException()) {
System.err.println("Error on process CAS call to remote service:");
List<Exception> exceptions = aStatus.getExceptions();
for (int i = 0; i < exceptions.size(); i++) {
((Throwable) exceptions.get(i)).printStackTrace();
}
}
try {
JCas cas = aCas.getJCas();
for(Token token : JCasUtil.select(cas, Token.class)) {
System.out.println(token.getCoveredText() + " " + token.getPos().getPosValue());
}
} catch (CASException e) {
e.printStackTrace();
}
}
}
/**
* CasConsumer would use tags and features to write output file, evaluate and print precision,
* recall and F-1 measure.
*
* @param arg0
* @throws ResourceProcessException
*/
@Override
public void processCas(CAS arg0) throws ResourceProcessException {
/** convert type of arg0 */
JCas jcas = null;
try {
jcas = arg0.getJCas();
} catch (CASException e1) {
// TODO Auto-generated catch block
e1.printStackTrace();
}
// TODO Auto-generated method stub
FSIterator<Annotation> ite = jcas.getAnnotationIndex(WordTag.type).iterator();
while (ite.hasNext()) {
/** collect features */
String id = ((WordTag) ite.get()).getId();
int begin = ((WordTag) ite.get()).getBegin0();
int end = ((WordTag) ite.get()).getEnd0();
String name = ((WordTag) ite.get()).getName();
/** organize string for output */
report.append(id);
report.append("|");
report.append(begin);
report.append(" ");
report.append(end);
report.append("|");
report.append(name);
report.append("\n");
/** count the length of output string */
count++;
ite.next();
}
result = report.toString();
File sampleOut = new File("src/main/resources/data/sample.out");
try {
testRecall = FileUtils.file2String(sampleOut);
} catch (IOException e1) {
// TODO Auto-generated catch block
e1.printStackTrace();
}
/** split strings from file into sentences */
String[] resultSplit = result.split("\n");
String[] recallSplit = testRecall.split("\n");
PrecisionRecallCalculator(recallSplit, resultSplit);
/** write the output file to the project root */
String path = "hw1-longh.out";
File dirFile = new File(path);
/** make sure no conflict */
if (dirFile.exists()) {
dirFile.delete();
}
try {
/** write file */
BufferedWriter bw1 = new BufferedWriter(new FileWriter(path, true));
bw1.write(report.toString());
bw1.flush();
bw1.close();
} catch (IOException e) {
e.printStackTrace();
}
}
/* (non-Javadoc)
* @see org.apache.uima.collection.base_cpm.CasObjectProcessor#processCas(org.apache.uima.cas.CAS)
*/
public void processCas(CAS aCAS) throws ResourceProcessException {
JCas jcas;
try {
jcas = aCAS.getJCas();
} catch (CASException e) {
logger.log(Level.SEVERE, e.getMessage());
throw new ResourceProcessException(e);
}
TweetAnnotation tweetAnn =
(TweetAnnotation) jcas.getAnnotationIndex(TweetAnnotation.type).iterator().next();
OMTweet answerTweet = evalCorpusReader.next();
if (!answerTweet.getId().equals(tweetAnn.getId())) {
logger.log(
Level.SEVERE,
"target corpus and evaluation corpus don't match to each other - "
+ answerTweet.getId()
+ ", "
+ tweetAnn.getId());
throw new ResourceProcessException();
}
String[] entity = extractEntityTags(answerTweet.getText());
String classified = null;
String prevClassified = null;
StringBuffer sb = new StringBuffer();
try {
sb.append("\n[");
sb.append(answerTweet.getPolarityString());
sb.append("=>");
sb.append(tweetAnn.getPolarity());
sb.append("] ");
sb.append(tweetAnn.getCoveredText());
sb.append('\n');
FSIterator<Annotation> tokenAnnIter =
jcas.getAnnotationIndex(TokenAnnotation.type).iterator();
TokenAnnotation tokenAnn = null;
int i = 0;
int prevClassifiedIdx = labelNoneIdx;
int prevAnswerIdx = labelNoneIdx;
String classifiedEntityStr = "";
String answerEntityStr = "";
while (tokenAnnIter.hasNext()) {
tokenAnn = (TokenAnnotation) tokenAnnIter.next();
classified = tokenAnn.getEntityLabel();
String answer = entity[i];
boolean correct = false;
if (classified.equals(answer)) {
correct = true;
}
int classifiedIdx = 0;
int answerIdx = 0;
try {
answerIdx = map.get(answer);
} catch (Exception e) {
logger.log(
Level.SEVERE,
"wrong annotation on the evaluation corpus - tweet id: "
+ answerTweet.getId()
+ ", answerTag="
+ answer);
logger.log(Level.SEVERE, e.getMessage());
answerIdx = map.get(labelNone);
}
try {
classifiedIdx = map.get(classified);
} catch (Exception e) {
logger.log(
Level.SEVERE,
"wrong annotation from the NER - tweet id: "
+ answerTweet.getId()
+ ", classifiedTag="
+ classified);
logger.log(Level.SEVERE, e.getMessage());
classifiedIdx = map.get(labelNone);
}
stat[classifiedIdx][0]++;
stat[answerIdx][1]++;
if (correct) {
stat[classifiedIdx][2]++;
}
if (classifiedIdx != labelNoneIdx) {
if (classifiedIdx / 3 != prevClassifiedIdx / 3) {
classifiedEntityCnt[classifiedIdx / 3]++;
if (prevClassifiedIdx != labelNoneIdx) {
sb.append('\t');
sb.append(classifiedEntityStr);
sb.append(" -> ");
sb.append(prevClassified.substring(0, prevClassified.lastIndexOf('_')));
sb.append('\n');
}
classifiedEntityStr = tokenAnn.getCoveredText();
} else {
classifiedEntityStr += " " + tokenAnn.getCoveredText();
}
} else if (prevClassifiedIdx != labelNoneIdx) {
sb.append('\t');
sb.append(classifiedEntityStr);
sb.append(" -> ");
sb.append(prevClassified.substring(0, prevClassified.lastIndexOf('_')));
sb.append('\n');
classifiedEntityStr = "";
}
prevClassifiedIdx = classifiedIdx;
if (answerIdx != labelNoneIdx) {
if (answerIdx / 3 != prevAnswerIdx / 3) {
answerEntityCnt[answerIdx / 3]++;
answerEntityStr = tokenAnn.getCoveredText();
} else {
answerEntityStr += " " + tokenAnn.getCoveredText();
}
} else if (prevAnswerIdx != labelNoneIdx) {
answerEntityStr = "";
}
prevAnswerIdx = answerIdx;
prevClassified = classified;
i++;
}
if (prevClassifiedIdx != labelNoneIdx) {
sb.append('\t');
sb.append(classifiedEntityStr);
sb.append(" -> ");
sb.append(prevClassified.substring(0, prevClassified.lastIndexOf('_')));
sb.append('\n');
}
// senti
String answerSenti = answerTweet.getPolarityString();
boolean correct = false;
String classifiedSenti = tweetAnn.getPolarity();
if (classifiedSenti.equals(senti)) {
correct = true;
}
int classifiedIdx = sentiIdx(classifiedSenti);
int answerIdx = sentiIdx(answerSenti);
senti[classifiedIdx][0]++;
senti[answerIdx][1]++;
if (classifiedIdx == answerIdx) {
correct = true;
}
if (correct) {
senti[classifiedIdx][2]++;
}
cnt++;
logger.log(Level.INFO, sb.toString());
} catch (CASRuntimeException e) {
throw new ResourceProcessException(e);
}
}
