Examples with Morpheme zemberek.morphology.morphotactics.Morpheme used on opensource projects

Example 1 with Morpheme

use of zemberek.morphology.morphotactics.Morpheme in project zemberek-nlp by ahmetaa.

the class WordGenerator method generate.

private List<Result> generate(String input, List<StemTransition> candidates, List<Morpheme> morphemes) {
    if (debugMode) {
        debugData = new AnalysisDebugData();
        debugData.input = input;
        debugData.candidateStemTransitions.addAll(candidates);
    }
    // generate initial search paths.
    List<GenerationPath> paths = new ArrayList<>();
    for (StemTransition candidate : candidates) {
        // we set the tail as " " because in morphotactics, some conditions look for tail's size
        // during graph walk. Because this is generation we let that condition pass always.
        SearchPath searchPath = SearchPath.initialPath(candidate, " ");
        List<Morpheme> morphemesInPath;
        // we skip it if it matches with the initial morpheme of the graph visiting SearchPath object.
        if (morphemes.size() > 0) {
            if (morphemes.get(0).equals(searchPath.getCurrentState().morpheme)) {
                morphemesInPath = morphemes.subList(1, morphemes.size());
            } else {
                morphemesInPath = new ArrayList<>(morphemes);
            }
        } else {
            morphemesInPath = new ArrayList<>(0);
        }
        paths.add(new GenerationPath(searchPath, morphemesInPath));
    }
    // search graph.
    List<GenerationPath> resultPaths = search(paths);
    // generate results from successful paths.
    List<Result> result = new ArrayList<>(resultPaths.size());
    for (GenerationPath path : resultPaths) {
        SingleAnalysis analysis = SingleAnalysis.fromSearchPath(path.path);
        result.add(new Result(analysis.surfaceForm(), analysis));
        if (debugMode) {
            debugData.results.add(analysis);
        }
    }
    return result;
}

Example 2 with Morpheme

use of zemberek.morphology.morphotactics.Morpheme in project zemberek-nlp by ahmetaa.

the class SingleAnalysis method fromSearchPath.

// Here we generate a SingleAnalysis from a search path.
public static SingleAnalysis fromSearchPath(SearchPath searchPath) {
    List<MorphemeData> morphemes = new ArrayList<>(searchPath.transitions.size());
    int derivationCount = 0;
    for (SurfaceTransition transition : searchPath.getTransitions()) {
        if (transition.isDerivative()) {
            derivationCount++;
        }
        Morpheme morpheme = transition.getMorpheme();
        // we skip these two morphemes as they create visual noise and does not carry much information.
        if (morpheme == TurkishMorphotactics.nom || morpheme == TurkishMorphotactics.pnon) {
            continue;
        }
        // if empty, use the cache.
        if (transition.surface.isEmpty()) {
            MorphemeData morphemeData = emptyMorphemeCache.get(morpheme);
            if (morphemeData == null) {
                morphemeData = new MorphemeData(morpheme, "");
                emptyMorphemeCache.put(morpheme, morphemeData);
            }
            morphemes.add(morphemeData);
            continue;
        }
        MorphemeData suffixSurface = new MorphemeData(morpheme, transition.surface);
        morphemes.add(suffixSurface);
    }
    int[] groupBoundaries = new int[derivationCount + 1];
    // we assume there is always an IG
    groupBoundaries[0] = 0;
    int morphemeCounter = 0, derivationCounter = 1;
    for (MorphemeData morphemeData : morphemes) {
        if (morphemeData.morpheme.derivational) {
            groupBoundaries[derivationCounter] = morphemeCounter;
            derivationCounter++;
        }
        morphemeCounter++;
    }
    // if dictionary item is `Dummy`, use the referenced item.
    // `Dummy` items are usually generated for some compound words. For example for `zeytinyağı`
    // a DictionaryItem is generated with root "zeytinyağ". But here we switch to the original.
    DictionaryItem item = searchPath.getDictionaryItem();
    if (item.hasAttribute(RootAttribute.Dummy)) {
        item = item.getReferenceItem();
    }
    return new SingleAnalysis(item, morphemes, groupBoundaries);
}

Example 3 with Morpheme

use of zemberek.morphology.morphotactics.Morpheme in project zemberek-nlp by ahmetaa.

the class WordGenerator method generate.

public List<Result> generate(String stem, List<String> morphemeIds) {
    List<Morpheme> morphemes = new ArrayList<>();
    for (String morphemeId : morphemeIds) {
        Morpheme morpheme = TurkishMorphotactics.getMorpheme(morphemeId);
        morphemes.add(morpheme);
    }
    List<StemTransition> candidates = stemTransitions.getPrefixMatches(stem, false);
    return generate(stem, candidates, morphemes);
}