예제 #1
0
 // look up and apply coarts for given rels to each sign in result
 private void applyCoarts(List<String> coartRels, Collection<Sign> result) {
   List<Sign> inputSigns = new ArrayList<Sign>(result);
   result.clear();
   List<Sign> outputSigns = new ArrayList<Sign>(inputSigns.size());
   // for each rel, lookup coarts and apply to input signs, storing results in output signs
   for (Iterator<String> it = coartRels.iterator(); it.hasNext(); ) {
     String rel = it.next();
     Collection<String> preds = (Collection<String>) _coartRelsToPreds.get(rel);
     if (preds == null) continue; // not expected
     Collection<Sign> coartResult = getSignsFromRelAndPreds(rel, preds);
     if (coartResult == null) continue;
     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);
 }
예제 #2
0
 /**
  * For a string of 1 or more surface words, return all of the lexical entries for each word as a
  * list of sign hashes. Tokenization is performed using the configured tokenizer.
  *
  * @param w the words in string format
  * @return a list of sign hashes
  * @exception LexException thrown if word not found
  */
 public List<SignHash> getEntriesFromWords(String s) throws LexException {
   List<SignHash> entries = new ArrayList<SignHash>();
   List<Word> words = tokenizer.tokenize(s);
   for (Iterator<Word> it = words.iterator(); it.hasNext(); ) {
     Word w = it.next();
     SignHash signs = getSignsFromWord(w);
     if (signs.size() == 0) {
       throw new LexException("Word not in lexicon: \"" + w + "\"");
     }
     entries.add(signs);
   }
   return entries;
 }
예제 #3
0
  /** Loads the lexicon and morph files. */
  public void init(URL lexiconUrl, URL morphUrl) throws IOException {

    List<Family> lexicon = null;
    List<MorphItem> morph = null;
    List<MacroItem> macroModel = null;

    // load category families (lexicon), morph forms and macros
    lexicon = getLexicon(lexiconUrl);
    Pair<List<MorphItem>, List<MacroItem>> morphInfo = getMorph(morphUrl);
    morph = morphInfo.a;
    macroModel = morphInfo.b;

    // index words; also index stems to words, as default preds
    // store indexed coarticulation attrs too
    _words = new GroupMap<Word, MorphItem>();
    _predToWords = new GroupMap<String, Word>();
    _coartAttrs = new HashSet<String>();
    _indexedCoartAttrs = new HashSet<String>();
    for (MorphItem morphItem : morph) {
      Word surfaceWord = morphItem.getSurfaceWord();
      _words.put(surfaceWord, morphItem);
      _predToWords.put(morphItem.getWord().getStem(), surfaceWord);
      if (morphItem.isCoart()) {
        Word indexingWord = morphItem.getCoartIndexingWord();
        _words.put(indexingWord, morphItem);
        Pair<String, String> first = indexingWord.getSurfaceAttrValPairs().next();
        _indexedCoartAttrs.add(first.a);
        for (Iterator<Pair<String, String>> it = surfaceWord.getSurfaceAttrValPairs();
            it.hasNext(); ) {
          Pair<String, String> p = it.next();
          _coartAttrs.add(p.a);
        }
      }
    }

    // index entries based on stem+pos
    _stems = new GroupMap<String, Object>();
    _posToEntries = new GroupMap<String, EntriesItem[]>();
    // index entries by supertag+pos, for supertagging
    _stagToEntries = new GroupMap<String, EntriesItem>();
    // also index rels and coart rels to preds
    _relsToPreds = new GroupMap<String, String>();
    _coartRelsToPreds = new GroupMap<String, String>();
    // and gather list of attributes used per atomic category type
    _catsToAttrs = new GroupMap<String, String>();
    _lfAttrs = new HashSet<String>();
    // and remember family and ent, names, for checking excluded list on morph items
    HashSet<String> familyAndEntryNames = new HashSet<String>();

    // index each family
    for (Family family : lexicon) {

      familyAndEntryNames.add(family.getName());
      EntriesItem[] entries = family.getEntries();
      DataItem[] data = family.getData();

      // for generic use when we get an unknown stem
      // from the morphological analyzer
      if (!family.isClosed()) {
        _posToEntries.put(family.getPOS(), entries);
      }

      // scan through entries
      for (int j = 0; j < entries.length; j++) {
        // index
        EntriesItem eItem = entries[j];
        _stagToEntries.put(eItem.getSupertag() + family.getPOS(), eItem);
        if (eItem.getStem().length() > 0) {
          _stems.put(eItem.getStem() + family.getPOS(), eItem);
        }
        try {
          // gather features
          eItem.getCat().forall(gatherAttrs);
          // record names
          familyAndEntryNames.add(eItem.getName());
          familyAndEntryNames.add(eItem.getQualifiedName());
        } catch (RuntimeException exc) {
          System.err.println("exception for: " + family.getName() + ": " + exc);
        }
      }

      // scan through data
      for (int j = 0; j < data.length; j++) {
        DataItem dItem = data[j];
        _stems.put(
            dItem.getStem() + family.getPOS(), new Pair<DataItem, EntriesItem[]>(dItem, entries));
        // index non-default preds to words
        if (!dItem.getStem().equals(dItem.getPred())) {
          Collection<Word> words = (Collection<Word>) _predToWords.get(dItem.getStem());
          if (words == null) {
            if (!openlex) {
              System.out.print("Warning: couldn't find words for pred '");
              System.out.println(dItem.getPred() + "' with stem '" + dItem.getStem() + "'");
            }
          } else {
            for (Iterator<Word> it = words.iterator(); it.hasNext(); ) {
              _predToWords.put(dItem.getPred(), it.next());
            }
          }
        }
      }

      // index rels to preds
      // nb: this covers relational (eg @x<GenRel>e) and featural (eg @e<tense>past)
      //     elementary predications
      List<String> indexRels = new ArrayList<String>(3);
      String familyIndexRel = family.getIndexRel();
      if (familyIndexRel.length() > 0) {
        indexRels.add(familyIndexRel);
      }
      for (int j = 0; j < entries.length; j++) {
        EntriesItem eItem = entries[j];
        String indexRel = eItem.getIndexRel();
        if (indexRel.length() > 0 && !indexRel.equals(familyIndexRel)) {
          indexRels.add(indexRel);
        }
      }
      for (Iterator<String> it = indexRels.iterator(); it.hasNext(); ) {
        String indexRel = it.next();
        // nb: not indexing on entries items, b/c some stems are still defaults
        for (int j = 0; j < data.length; j++) {
          DataItem dItem = data[j];
          _relsToPreds.put(indexRel, dItem.getPred());
        }
      }

      // index coart rels (features, really) to preds
      String coartRel = family.getCoartRel();
      if (coartRel.length() > 0) {
        for (int j = 0; j < data.length; j++) {
          _coartRelsToPreds.put(coartRel, data[j].getPred());
        }
      }
    }

    // index the macros
    _macros = new GroupMap<String, FeatureStructure>();
    // nb: could just index MacroItem objects for feature structures too;
    //     this might be a bit cleaner, but life is short
    _macroItems = new HashMap<String, MacroItem>();
    for (MacroItem mi : macroModel) {
      String macName = mi.getName();
      FeatureStructure[] specs = mi.getFeatureStructures();
      for (int j = 0; j < specs.length; j++) {
        _macros.put(macName, specs[j]);
      }
      // this is for handling LF part of macros
      _macroItems.put(macName, mi);
    }

    // with morph items, check POS, macro names, excluded list for xref
    for (MorphItem morphItem : morph) {
      Word w = morphItem.getWord();
      if (!openlex
          && !_stems.containsKey(w.getStem() + w.getPOS())
          && !_posToEntries.containsKey(w.getPOS())) {
        System.err.println(
            "Warning: no entries for stem '"
                + w.getStem()
                + "' and POS '"
                + w.getPOS()
                + "' found for word '"
                + w
                + "'");
      }
      String[] macroNames = morphItem.getMacros();
      for (int j = 0; j < macroNames.length; j++) {
        if (!_macroItems.containsKey(macroNames[j])) {
          System.err.println(
              "Warning: macro "
                  + macroNames[j]
                  + " not found for word '"
                  + morphItem.getWord()
                  + "'");
        }
      }
      String[] excludedNames = morphItem.getExcluded();
      for (int j = 0; j < excludedNames.length; j++) {
        if (!familyAndEntryNames.contains(excludedNames[j])) {
          System.err.println(
              "Warning: excluded family or entry '"
                  + excludedNames[j]
                  + "' not found for word '"
                  + morphItem.getWord()
                  + "'");
        }
      }
    }
  }