/** 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()
+ "'");
}
}
}
}
// given MorphItem
private void getWithMorphItem(
Word w, MorphItem mi, String targetPred, String targetRel, SignHash result)
throws LexException {
// get supertags for filtering, if a supertagger is installed
Map<String, Double> supertags = null;
Set<String> supertagsFound = null;
if (_supertagger != null) {
supertags = _supertagger.getSupertags();
if (supertags != null) supertagsFound = new HashSet<String>(supertags.size());
}
// get macro adder
MacroAdder macAdder = getMacAdder(mi);
// if we have this stem in our lexicon
String stem = mi.getWord().getStem();
String pos = mi.getWord().getPOS();
Set<EntriesItem[]> explicitEntries =
null; // for storing entries from explicitly listed family members
if (_stems.containsKey(stem + pos)) {
explicitEntries = new HashSet<EntriesItem[]>();
Collection<Object> stemItems = (Collection<Object>) _stems.get(stem + pos);
for (Iterator<Object> I = stemItems.iterator(); I.hasNext(); ) {
Object item = I.next();
// see if it's an EntriesItem
if (item instanceof EntriesItem) {
EntriesItem entry = (EntriesItem) item;
// do lookup
getWithEntriesItem(
w,
mi,
stem,
stem,
targetPred,
targetRel,
entry,
macAdder,
supertags,
supertagsFound,
result);
}
// otherwise it has to be a Pair containing a DataItem and
// an EntriesItem[]
else {
@SuppressWarnings("rawtypes")
DataItem dItem = (DataItem) ((Pair) item).a;
@SuppressWarnings("rawtypes")
EntriesItem[] entries = (EntriesItem[]) ((Pair) item).b;
// store entries
explicitEntries.add(entries);
// do lookup
getWithDataItem(
w,
mi,
dItem,
entries,
targetPred,
targetRel,
macAdder,
supertags,
supertagsFound,
result);
}
}
}
// for entries that are not explicitly in the lexicon file, we have to create
// Signs from the open class entries with the appropriate part-of-speech
Collection<EntriesItem[]> entrySets = (Collection<EntriesItem[]>) _posToEntries.get(pos);
if (entrySets != null) {
for (Iterator<EntriesItem[]> E = entrySets.iterator(); E.hasNext(); ) {
EntriesItem[] entries = E.next();
// skip if entries explicitly listed
if (explicitEntries != null && explicitEntries.contains(entries)) continue;
// otherwise get entries with pred = targetPred, or stem if null
String pred = (targetPred != null) ? targetPred : stem;
getWithDataItem(
w,
mi,
new DataItem(stem, pred),
entries,
targetPred,
targetRel,
macAdder,
supertags,
supertagsFound,
result);
}
}
// finally do entries for any remaining supertags
if (supertags != null) {
for (String supertag : supertags.keySet()) {
if (supertagsFound.contains(supertag)) continue;
Set<EntriesItem> entries = _stagToEntries.get(supertag + pos);
if (entries == null) continue; // nb: could be a POS mismatch
// get entries with pred = targetPred, or stem if null
String pred = (targetPred != null) ? targetPred : stem;
for (EntriesItem entry : entries) {
if (!entry.getStem().equals(DEFAULT_VAL)) continue;
getWithEntriesItem(
w,
mi,
stem,
pred,
targetPred,
targetRel,
entry,
macAdder,
supertags,
supertagsFound,
result);
}
}
}
}
/**
* Returns whether the given rel (semantic feature, really) is one used to signal coarticulation.
*/
public boolean isCoartRel(String rel) {
return _coartRelsToPreds.containsKey(rel);
}