private boolean matchLemma(Token token, String primitiveMask) {
boolean match = false;
String[] lemmas = token.getPrimitive();
if (lemmas != null) {
for (String lemma : lemmas) {
if (lemma.equalsIgnoreCase(primitiveMask)) {
match = true;
break;
}
}
}
return match;
}
/**
* Applies all active rules described in Rules.xml given a sentence properly tokenized, tagged,
* chunked and shallow parsed.
*
* @param sentence a tokenized, tagged, chunked and shallow parsed sentence.
* @param dictionary a word and tag dictionary
* @return a list containing all the mistakes found in the sentence. Each mistake can be localized
* between the character indexes given in the span field of the mistake.
*/
public List<Mistake> check(Sentence sentence) {
long start = 0;
if (LOGGER.isDebugEnabled()) {
start = System.nanoTime();
}
// Insert two empty tokens at the sentence start and end
List<Token> tokens = new ArrayList<Token>();
Token empty1 = new TokenCogroo("", new Span(0, 0));
empty1.setMorphologicalTag(new MorphologicalTag());
tokens.add(empty1);
tokens.addAll(sentence.getTokens());
Token empty2 = new TokenCogroo("", new Span(0, 0));
empty2.setMorphologicalTag(new MorphologicalTag());
tokens.add(empty2);
sentence.setTokens(tokens);
// mistakes will hold mistakes found in the sentence.
List<Mistake> mistakes = new ArrayList<Mistake>();
// rules will hold the tree being used to seek for mistakes.
RulesTree rulesTree;
if (RulesProperties.APPLY_LOCAL) {
// Seeks for errors that can occur anywhere in the sentence (general).
rulesTree = this.rulesTreesProvider.getTrees().getGeneral();
// For each token in the sentence.
for (int i = 0; i < sentence.getTokens().size(); i++) {
// For each token, gets back to the initial state (hence 0).
List<State> nextStates = rulesTree.getRoot().getNextStates();
// i is the index of the token that began the rule applying process.
mistakes = this.getMistakes(mistakes, nextStates, sentence, i, i, sentence);
}
}
// remove aux tokens
sentence.setTokens(sentence.getTokens().subList(1, sentence.getTokens().size() - 1));
if (RulesProperties.APPLY_PHRASE_LOCAL) {
// Seeks for errors inside a chunk (phrase local).
rulesTree = this.rulesTreesProvider.getTrees().getPhraseLocal();
// For each chunk in the sentence.
List<Chunk> chunks = sentence.getChunks();
for (int i = 0; i < chunks.size(); i++) {
for (int j = 0; j < chunks.get(i).getTokens().size(); j++) {
// For each token, gets back to the initial state (hence 0).
List<State> nextStates = rulesTree.getRoot().getNextStates();
// j is the index of the token that began the rule applying process.
mistakes = this.getMistakes(mistakes, nextStates, chunks.get(i), j, j, sentence);
}
}
}
if (RulesProperties.APPLY_SUBJECT_VERB) {
// Seeks for errors between a subject and a main verb.
rulesTree = this.rulesTreesProvider.getTrees().getSubjectVerb();
// For each chunk in the sentence.
List<SyntacticChunk> syntacticChunks = sentence.getSyntacticChunks();
for (int i = 0; i < syntacticChunks.size(); i++) {
List<State> nextStates = rulesTree.getRoot().getNextStates();
mistakes = this.getMistakes(mistakes, nextStates, syntacticChunks, i, i, sentence);
}
}
if (LOGGER.isDebugEnabled()) {
LOGGER.debug("Rules applied in " + (System.nanoTime() - start) / 1000 + "us");
}
filterIgnoredRules(mistakes);
return mistakes;
}
/**
* Determines if a token is matched by a rule element.
*
* @param token the token to be matched by the element
* @param element the element to be matched against the token
* @return <code>true</code> if there's a match, <code>false</code> otherwise
*/
private boolean match(Token token, Element element, int baseTokenIndex, Sentence sentence) {
boolean match;
boolean negated;
// Sees if the mask must or not match.
// Negated is optional, so it can be null, true or false.
// If null, consider as false.
if (element.isNegated() == null) {
match = false;
negated = false;
} else {
match = element.isNegated().booleanValue();
negated = element.isNegated().booleanValue();
}
for (Mask mask : element.getMask()) {
// If the token must match the mask.
if (!negated) {
// If not negated, match starts as false and just one match is needed to make it true.
if (mask.getLexemeMask() != null
&& mask.getLexemeMask().equalsIgnoreCase(token.getLexeme())) {
match = true;
} else if (mask.getPrimitiveMask() != null && matchLemma(token, mask.getPrimitiveMask())) {
match = true;
} else if (mask.getTagMask() != null && token.getMorphologicalTag() != null) {
match = match | token.getMorphologicalTag().matchExact(mask.getTagMask(), false);
} else if (mask.getTagReference() != null && token.getMorphologicalTag() != null) {
match =
match
| token
.getMorphologicalTag()
.match(
RuleUtils.createTagMaskFromReference(
mask.getTagReference(), sentence, baseTokenIndex),
false);
} else if (mask.getOutOfBounds() != null
&& (baseTokenIndex == 0 || baseTokenIndex == sentence.getTokens().size() - 1)) {
match = false;
}
} else { // The token must NOT match the mask.
// If negated, match starts as true and just one match is needed to make it false.
if (mask.getLexemeMask() != null
&& mask.getLexemeMask().equalsIgnoreCase(token.getLexeme())) {
match = false;
} else if (mask.getPrimitiveMask() != null && matchLemma(token, mask.getPrimitiveMask())) {
match = false;
} else if (mask.getTagMask() != null
&& token != null
&& token.getMorphologicalTag() != null) {
match = match & !token.getMorphologicalTag().matchExact(mask.getTagMask(), false);
} else if (mask.getTagReference() != null
&& token != null
&& token.getMorphologicalTag() != null) {
match =
match
& !token
.getMorphologicalTag()
.match(
RuleUtils.createTagMaskFromReference(
mask.getTagReference(), sentence, baseTokenIndex),
false);
} else if (mask.getOutOfBounds() != null
&& (baseTokenIndex == 0 || baseTokenIndex == sentence.getTokens().size() - 1)) {
match = false;
}
}
}
return match;
}
