// look up and apply coarts for w to each sign in result @SuppressWarnings("unchecked") private void applyCoarts(Word w, SignHash result) throws LexException { List<Sign> inputSigns = new ArrayList<Sign>(result.asSignSet()); result.clear(); List<Sign> outputSigns = new ArrayList<Sign>(inputSigns.size()); // for each surface attr, lookup coarts and apply to input signs, storing results in output // signs for (Iterator<Pair<String, String>> it = w.getSurfaceAttrValPairs(); it.hasNext(); ) { Pair<String, String> p = it.next(); String attr = (String) p.a; if (!_indexedCoartAttrs.contains(attr)) continue; String val = (String) p.b; Word coartWord = Word.createWord(attr, val); SignHash coartResult = getSignsFromWord(coartWord, null, null, null); for (Iterator<Sign> it2 = coartResult.iterator(); it2.hasNext(); ) { Sign coartSign = it2.next(); // apply to each input for (int j = 0; j < inputSigns.size(); j++) { Sign sign = inputSigns.get(j); grammar.rules.applyCoart(sign, coartSign, outputSigns); } } // switch output to input for next iteration inputSigns.clear(); inputSigns.addAll(outputSigns); outputSigns.clear(); } // add results back result.addAll(inputSigns); }
// get signs using an additional arg for a target rel private Collection<Sign> getSignsFromPredAndTargetRel(String pred, String targetRel) { Collection<Word> words = (Collection<Word>) _predToWords.get(pred); String specialTokenConst = null; // for robustness, when using supertagger, add words for pred sans sense index int dotIndex = -1; if (_supertagger != null && !Character.isDigit(pred.charAt(0)) && // skip numbers (dotIndex = pred.lastIndexOf('.')) > 0 && pred.length() > dotIndex + 1 && pred.charAt(dotIndex + 1) != '_') // skip titles, eg Mr._Smith { String barePred = pred.substring(0, dotIndex); Collection<Word> barePredWords = (Collection<Word>) _predToWords.get(barePred); if (words == null) words = barePredWords; else if (barePredWords != null) { Set<Word> unionWords = new HashSet<Word>(words); unionWords.addAll(barePredWords); words = unionWords; } } if (words == null) { specialTokenConst = tokenizer.getSpecialTokenConstant(tokenizer.isSpecialToken(pred)); if (specialTokenConst == null) return null; // lookup words with pred = special token const Collection<Word> specialTokenWords = (Collection<Word>) _predToWords.get(specialTokenConst); // replace special token const with pred if (specialTokenWords == null) return null; words = new ArrayList<Word>(specialTokenWords.size()); for (Iterator<Word> it = specialTokenWords.iterator(); it.hasNext(); ) { Word stw = it.next(); Word w = Word.createSurfaceWord(stw, pred); words.add(w); } } List<Sign> retval = new ArrayList<Sign>(); for (Iterator<Word> it = words.iterator(); it.hasNext(); ) { Word w = it.next(); try { SignHash signs = getSignsFromWord(w, specialTokenConst, pred, targetRel); retval.addAll(signs.asSignSet()); } // shouldn't happen catch (LexException exc) { System.err.println("Unexpected lex exception for word " + w + ": " + exc); } } return retval; }