private static void updateResultWithActualReferents(
Result result,
List<AnaphorWithReferent> anaphorsWithReferents,
int sentenceId,
int pronominalIndex,
Coref inputPronominal) {
List<AnaphorWithReferent> anaphorsForSentence =
getAnaphorsForSentence(anaphorsWithReferents, sentenceId);
AnaphorWithReferent anaphorByTokenIndex =
getAnaphorByTokenIndex(anaphorsForSentence, pronominalIndex);
if (anaphorByTokenIndex == null) {
result.setActualReferentSentenceId(-1);
result.setActualReferentStartIndex(-1);
result.setActualReferentText(null);
return;
}
String actualPronominalText = anaphorByTokenIndex.getAnaphor().toString();
if (!validateActualPronominal(inputPronominal.toString(), actualPronominalText)) {
String errorMessage =
String.format(
"Actual pronominal wasn't found for: sentence %s token %s",
result.getSentenceId(), result.getExpectedReferentStartIndex());
System.err.println(errorMessage);
return;
}
NpTreeInSentence actualReferent = anaphorByTokenIndex.getReferent();
result.setActualReferentSentenceId(actualReferent.getSentence().getId());
result.setActualReferentText(actualReferent.toString());
result.setActualReferentStartIndex(actualReferent.getStartIndex());
}
private static List<Result> compareWithInputText(
InputText inputText, List<AnaphorWithReferent> anaphorsWithReferents) {
List<Result> results = new ArrayList<>();
for (InputSentence inputSentence : inputText.getSentences()) {
if (!inputSentence.isValid()) {
continue;
}
List<Coref> pronominalCorefs = inputSentence.getPronominalCorefs();
Result result;
if (pronominalCorefs.size() == 0) {
result = new Result(inputText.getId(), inputSentence.getSentenceId());
result.setSentenceText(inputText.getSentenceById(inputSentence.getSentenceId()).toString());
result.setExpectedReferentSentenceId(null);
result.setExpectedReferentStartIndex(null);
result.setExpectedReferentText(null);
// Assuming only one
List<AnaphorWithReferent> anaphorsForSentence =
getAnaphorsForSentence(anaphorsWithReferents, inputSentence.getSentenceId());
if (anaphorsForSentence.size() == 0) {
continue;
}
AnaphorWithReferent anaphorForSentence = anaphorsForSentence.get(0);
// TODO: check if the same anaphor
NpTreeInSentence referent = anaphorForSentence.getReferent();
result.setActualReferentSentenceId(referent.getSentence().getId());
result.setActualReferentText(referent.toString());
result.setActualReferentStartIndex(referent.getStartIndex());
results.add(result);
continue;
}
for (Coref inputPronominal : pronominalCorefs) {
result = new Result(inputText.getId(), inputSentence.getSentenceId());
result.setSentenceText(inputText.getSentenceById(inputSentence.getSentenceId()).toString());
int sentenceId = inputPronominal.getSentenceId();
int pronominalIndex = inputPronominal.getStartIndex();
result.setPronoun(inputPronominal.toString());
result.setPronounTokenIndex(pronominalIndex);
List<Coref> expectedReferentsForPronominal =
inputText.getReferentsForPronominal(inputPronominal.getId());
if (expectedReferentsForPronominal.size() >= 1) {
// We take the first referent from the sentence.
List<String> corefs =
expectedReferentsForPronominal
.stream()
.map(c -> c.toString())
.collect(Collectors.toList());
List<Integer> sentenceIndices =
expectedReferentsForPronominal
.stream()
.map(c -> c.getSentenceId())
.collect(Collectors.toList());
List<Integer> startIndices =
expectedReferentsForPronominal
.stream()
.map(c -> c.getStartIndex())
.collect(Collectors.toList());
result.setExpectedReferentSentenceId(sentenceIndices);
result.setExpectedReferentText(corefs);
result.setExpectedReferentStartIndex(startIndices);
} else {
result.setExpectedReferentSentenceId(null);
result.setExpectedReferentText(null);
result.setExpectedReferentStartIndex(null);
}
updateResultWithActualReferents(
result, anaphorsWithReferents, sentenceId, pronominalIndex, inputPronominal);
results.add(result);
}
}
return results;
}