Examples with TurkicLetter zemberek.core.turkish.TurkicLetter used on opensource projects

Example 1 with TurkicLetter

use of zemberek.core.turkish.TurkicLetter in project zemberek-nlp by ahmetaa.

the class SuffixSurfaceNodeGenerator method generate.

public List<SuffixSurfaceNode> generate(EnumSet<PhoneticAttribute> attrs, EnumSet<PhoneticExpectation> expectations, SuffixData suffixData, SuffixForm suffixForm) {
    List<SuffixToken> tokenList = Lists.newArrayList(new SuffixStringTokenizer(suffixForm.generation));
    // zero length token
    if (tokenList.size() == 0) {
        return Lists.newArrayList(new SuffixSurfaceNode(suffixForm, "", attrs.clone(), expectations.clone(), suffixData, suffixForm.terminationType));
    }
    List<SuffixSurfaceNode> forms = new ArrayList<SuffixSurfaceNode>(1);
    // generation of forms. normally only one form is generated. But in situations like cI~k, two Forms are generated.
    TurkishLetterSequence seq = new TurkishLetterSequence();
    int index = 0;
    for (SuffixToken token : tokenList) {
        EnumSet<PhoneticAttribute> formAttrs = defineMorphemicAttributes(seq, attrs);
        switch(token.type) {
            case LETTER:
                seq.append(token.letter);
                if (index == tokenList.size() - 1) {
                    forms.add(new SuffixSurfaceNode(suffixForm, seq.toString(), defineMorphemicAttributes(seq, attrs), suffixForm.terminationType));
                }
                break;
            case A_WOVEL:
                if (index == 0 && attrs.contains(LastLetterVowel)) {
                    break;
                }
                TurkicLetter lA = TurkicLetter.UNDEFINED;
                if (formAttrs.contains(LastVowelBack)) {
                    lA = L_a;
                } else if (formAttrs.contains(LastVowelFrontal)) {
                    lA = L_e;
                }
                if (lA == TurkicLetter.UNDEFINED) {
                    throw new IllegalArgumentException("Cannot generate A form!");
                }
                seq.append(lA);
                if (index == tokenList.size() - 1) {
                    forms.add(new SuffixSurfaceNode(suffixForm, seq.toString(), defineMorphemicAttributes(seq, attrs), suffixForm.terminationType));
                }
                break;
            case I_WOVEL:
                if (index == 0 && attrs.contains(LastLetterVowel)) {
                    break;
                }
                TurkicLetter li = TurkicLetter.UNDEFINED;
                if (formAttrs.containsAll(Arrays.asList(LastVowelBack, LastVowelRounded))) {
                    li = L_u;
                } else if (formAttrs.containsAll(Arrays.asList(LastVowelBack, LastVowelUnrounded))) {
                    li = L_ii;
                } else if (formAttrs.containsAll(Arrays.asList(LastVowelFrontal, LastVowelRounded))) {
                    li = L_uu;
                } else if (formAttrs.containsAll(Arrays.asList(LastVowelFrontal, LastVowelUnrounded))) {
                    li = L_i;
                }
                if (li == TurkicLetter.UNDEFINED) {
                    throw new IllegalArgumentException("Cannot generate I form!");
                }
                seq.append(li);
                if (index == tokenList.size() - 1) {
                    forms.add(new SuffixSurfaceNode(suffixForm, seq.toString(), defineMorphemicAttributes(seq, attrs), suffixForm.terminationType));
                }
                break;
            case APPEND:
                if (formAttrs.contains(LastLetterVowel)) {
                    seq.append(token.letter);
                }
                if (index == tokenList.size() - 1) {
                    forms.add(new SuffixSurfaceNode(suffixForm, seq.toString(), defineMorphemicAttributes(seq, attrs), suffixForm.terminationType));
                }
                break;
            case DEVOICE_FIRST:
                TurkicLetter ld = token.letter;
                if (formAttrs.contains(LastLetterVoiceless)) {
                    ld = Turkish.Alphabet.devoice(token.letter);
                }
                seq.append(ld);
                if (index == tokenList.size() - 1) {
                    forms.add(new SuffixSurfaceNode(suffixForm, seq.toString(), defineMorphemicAttributes(seq, attrs), suffixForm.terminationType));
                }
                break;
            case VOICE_LAST:
                ld = token.letter;
                seq.append(ld);
                if (index == tokenList.size() - 1) {
                    forms.add(new SuffixSurfaceNode(suffixForm, seq.toString(), defineMorphemicAttributes(seq, attrs), EnumSet.of(PhoneticExpectation.ConsonantStart), suffixData, suffixForm.terminationType));
                    seq.changeLast(Turkish.Alphabet.voice(token.letter));
                    forms.add(new SuffixSurfaceNode(suffixForm, seq.toString(), defineMorphemicAttributes(seq, attrs), EnumSet.of(PhoneticExpectation.VowelStart), suffixData, TerminationType.NON_TERMINAL));
                }
                break;
        }
        index++;
    }
    return forms;
}

Example 2 with TurkicLetter

use of zemberek.core.turkish.TurkicLetter in project zemberek-nlp by ahmetaa.

the class AttributesHelper method getMorphemicAttributes.

public static AttributeSet<PhoneticAttribute> getMorphemicAttributes(CharSequence seq, AttributeSet<PhoneticAttribute> predecessorAttrs) {
    if (seq.length() == 0) {
        return predecessorAttrs.copy();
    }
    AttributeSet<PhoneticAttribute> attrs = new AttributeSet<>();
    if (alphabet.containsVowel(seq)) {
        TurkicLetter last = alphabet.getLastLetter(seq);
        if (last.isVowel()) {
            attrs.add(LastLetterVowel);
        } else {
            attrs.add(LastLetterConsonant);
        }
        TurkicLetter lastVowel = last.isVowel() ? last : alphabet.getLastVowel(seq);
        if (lastVowel.isFrontal()) {
            attrs.add(LastVowelFrontal);
        } else {
            attrs.add(LastVowelBack);
        }
        if (lastVowel.isRounded()) {
            attrs.add(LastVowelRounded);
        } else {
            attrs.add(LastVowelUnrounded);
        }
        if (alphabet.getFirstLetter(seq).isVowel()) {
            attrs.add(FirstLetterVowel);
        } else {
            attrs.add(FirstLetterConsonant);
        }
    } else {
        // we transfer vowel attributes from the predecessor attributes.
        attrs.copyFrom(predecessorAttrs);
        attrs.addAll(NO_VOWEL_ATTRIBUTES);
        attrs.remove(LastLetterVowel);
        attrs.remove(ExpectsConsonant);
    }
    TurkicLetter last = alphabet.getLastLetter(seq);
    if (last.isVoiceless()) {
        attrs.add(LastLetterVoiceless);
        if (last.isStopConsonant()) {
            // kitap
            attrs.add(LastLetterVoicelessStop);
        }
    } else {
        attrs.add(LastLetterVoiced);
    }
    return attrs;
}

Example 3 with TurkicLetter

use of zemberek.core.turkish.TurkicLetter in project zemberek-nlp by ahmetaa.

the class StemTransitionGenerator method generateModifiedRootNodes.

private List<StemTransition> generateModifiedRootNodes(DictionaryItem dicItem) {
    StringBuilder modifiedSeq = new StringBuilder(dicItem.pronunciation);
    AttributeSet<PhoneticAttribute> originalAttrs = calculateAttributes(dicItem.pronunciation);
    AttributeSet<PhoneticAttribute> modifiedAttrs = originalAttrs.copy();
    MorphemeState modifiedRootState = null;
    MorphemeState unmodifiedRootState = null;
    for (RootAttribute attribute : dicItem.attributes) {
        // generate other boundary attributes and modified root state.
        switch(attribute) {
            case Voicing:
                char last = alphabet.getLastChar(modifiedSeq);
                char voiced = alphabet.voice(last);
                if (last == voiced) {
                    throw new LexiconException("Voicing letter is not proper in:" + dicItem);
                }
                if (dicItem.lemma.endsWith("nk")) {
                    voiced = 'g';
                }
                modifiedSeq.setCharAt(modifiedSeq.length() - 1, voiced);
                modifiedAttrs.remove(PhoneticAttribute.LastLetterVoicelessStop);
                originalAttrs.add(PhoneticAttribute.ExpectsConsonant);
                modifiedAttrs.add(PhoneticAttribute.ExpectsVowel);
                // TODO: find a better way for this.
                modifiedAttrs.add(PhoneticAttribute.CannotTerminate);
                break;
            case Doubling:
                modifiedSeq.append(alphabet.getLastChar(modifiedSeq));
                originalAttrs.add(PhoneticAttribute.ExpectsConsonant);
                modifiedAttrs.add(PhoneticAttribute.ExpectsVowel);
                modifiedAttrs.add(PhoneticAttribute.CannotTerminate);
                break;
            case LastVowelDrop:
                TurkicLetter lastLetter = alphabet.getLastLetter(modifiedSeq);
                if (lastLetter.isVowel()) {
                    modifiedSeq.deleteCharAt(modifiedSeq.length() - 1);
                    modifiedAttrs.add(PhoneticAttribute.ExpectsConsonant);
                    modifiedAttrs.add(PhoneticAttribute.CannotTerminate);
                } else {
                    modifiedSeq.deleteCharAt(modifiedSeq.length() - 2);
                    if (!dicItem.primaryPos.equals(PrimaryPos.Verb)) {
                        originalAttrs.add(PhoneticAttribute.ExpectsConsonant);
                    } else {
                        unmodifiedRootState = morphotactics.verbLastVowelDropUnmodRoot_S;
                        modifiedRootState = morphotactics.verbLastVowelDropModRoot_S;
                    }
                    modifiedAttrs.add(PhoneticAttribute.ExpectsVowel);
                    modifiedAttrs.add(PhoneticAttribute.CannotTerminate);
                }
                break;
            case InverseHarmony:
                originalAttrs.add(PhoneticAttribute.LastVowelFrontal);
                originalAttrs.remove(PhoneticAttribute.LastVowelBack);
                modifiedAttrs.add(PhoneticAttribute.LastVowelFrontal);
                modifiedAttrs.remove(PhoneticAttribute.LastVowelBack);
                break;
            case ProgressiveVowelDrop:
                modifiedSeq.deleteCharAt(modifiedSeq.length() - 1);
                if (alphabet.containsVowel(modifiedSeq)) {
                    modifiedAttrs = calculateAttributes(modifiedSeq);
                }
                modifiedAttrs.add(PhoneticAttribute.LastLetterDropped);
                break;
            default:
                break;
        }
    }
    if (unmodifiedRootState == null) {
        unmodifiedRootState = morphotactics.getRootState(dicItem, originalAttrs);
    }
    StemTransition original = new StemTransition(dicItem.root, dicItem, originalAttrs, unmodifiedRootState);
    // if modified root state is not defined in the switch block, get it from morphotactics.
    if (modifiedRootState == null) {
        modifiedRootState = morphotactics.getRootState(dicItem, modifiedAttrs);
    }
    StemTransition modified = new StemTransition(modifiedSeq.toString(), dicItem, modifiedAttrs, modifiedRootState);
    if (original.equals(modified)) {
        return Collections.singletonList(original);
    }
    return Lists.newArrayList(original, modified);
}

Example 4 with TurkicLetter

use of zemberek.core.turkish.TurkicLetter in project zemberek-nlp by ahmetaa.

the class StemNodeGenerator method generateModifiedRootNodes.

private StemNode[] generateModifiedRootNodes(DictionaryItem dicItem) {
    if (dicItem.hasAttribute(Special)) {
        return handleSpecialStems(dicItem);
    }
    TurkishLetterSequence modifiedSeq = new TurkishLetterSequence(dicItem.pronunciation, alphabet);
    EnumSet<PhoneticAttribute> originalAttrs = calculateAttributes(dicItem.pronunciation);
    EnumSet<PhoneticAttribute> modifiedAttrs = originalAttrs.clone();
    EnumSet<PhoneticExpectation> originalExpectations = EnumSet.noneOf(PhoneticExpectation.class);
    EnumSet<PhoneticExpectation> modifiedExpectations = EnumSet.noneOf(PhoneticExpectation.class);
    for (RootAttribute attribute : dicItem.attributes) {
        // generate other boundary attributes and modified root state.
        switch(attribute) {
            case Voicing:
                TurkicLetter last = modifiedSeq.lastLetter();
                TurkicLetter modifiedLetter = alphabet.voice(last);
                if (modifiedLetter == null) {
                    throw new LexiconException("Voicing letter is not proper in:" + dicItem);
                }
                if (dicItem.lemma.endsWith("nk")) {
                    modifiedLetter = TurkishAlphabet.L_g;
                }
                modifiedSeq.changeLetter(modifiedSeq.length() - 1, modifiedLetter);
                modifiedAttrs.remove(PhoneticAttribute.LastLetterVoicelessStop);
                originalExpectations.add(PhoneticExpectation.ConsonantStart);
                modifiedExpectations.add(PhoneticExpectation.VowelStart);
                break;
            case Doubling:
                modifiedSeq.append(modifiedSeq.lastLetter());
                originalExpectations.add(PhoneticExpectation.ConsonantStart);
                modifiedExpectations.add(PhoneticExpectation.VowelStart);
                break;
            case LastVowelDrop:
                if (modifiedSeq.lastLetter().isVowel()) {
                    modifiedSeq.delete(modifiedSeq.length() - 1);
                    modifiedExpectations.add(PhoneticExpectation.ConsonantStart);
                } else {
                    modifiedSeq.delete(modifiedSeq.length() - 2);
                    if (!dicItem.primaryPos.equals(PrimaryPos.Verb)) {
                        originalExpectations.add(PhoneticExpectation.ConsonantStart);
                    }
                    modifiedExpectations.add(PhoneticExpectation.VowelStart);
                }
                break;
            case InverseHarmony:
                originalAttrs.add(PhoneticAttribute.LastVowelFrontal);
                originalAttrs.remove(PhoneticAttribute.LastVowelBack);
                modifiedAttrs.add(PhoneticAttribute.LastVowelFrontal);
                modifiedAttrs.remove(PhoneticAttribute.LastVowelBack);
                break;
            case ProgressiveVowelDrop:
                modifiedSeq.delete(modifiedSeq.length() - 1);
                if (modifiedSeq.hasVowel()) {
                    modifiedAttrs = calculateAttributes(modifiedSeq);
                }
                break;
            default:
                break;
        }
    }
    StemNode original = new StemNode(dicItem.root, dicItem, originalAttrs, originalExpectations);
    StemNode modified = new StemNode(modifiedSeq.toString(), dicItem, modifiedAttrs, modifiedExpectations);
    SuffixData[] roots = suffixProvider.defineSuccessorSuffixes(dicItem);
    original.exclusiveSuffixData = roots[0];
    modified.exclusiveSuffixData = roots[1];
    if (original.equals(modified)) {
        return new StemNode[] { original };
    }
    modified.setTermination(TerminationType.NON_TERMINAL);
    if (dicItem.hasAttribute(RootAttribute.CompoundP3sgRoot)) {
        original.setTermination(TerminationType.NON_TERMINAL);
    }
    return new StemNode[] { original, modified };
}

Example 5 with TurkicLetter

use of zemberek.core.turkish.TurkicLetter in project zemberek-nlp by ahmetaa.

the class StemTransitionsBase method generateModifiedRootNodes.

private List<StemTransition> generateModifiedRootNodes(DictionaryItem dicItem) {
    StringBuilder modifiedSeq = new StringBuilder(dicItem.pronunciation);
    AttributeSet<PhoneticAttribute> originalAttrs = calculateAttributes(dicItem.pronunciation);
    AttributeSet<PhoneticAttribute> modifiedAttrs = originalAttrs.copy();
    MorphemeState modifiedRootState = null;
    MorphemeState unmodifiedRootState = null;
    for (RootAttribute attribute : dicItem.attributes) {
        // generate other boundary attributes and modified root state.
        switch(attribute) {
            case Voicing:
                char last = alphabet.lastChar(modifiedSeq);
                char voiced = alphabet.voice(last);
                if (last == voiced) {
                    throw new LexiconException("Voicing letter is not proper in:" + dicItem);
                }
                if (dicItem.lemma.endsWith("nk")) {
                    voiced = 'g';
                }
                modifiedSeq.setCharAt(modifiedSeq.length() - 1, voiced);
                modifiedAttrs.remove(PhoneticAttribute.LastLetterVoicelessStop);
                originalAttrs.add(PhoneticAttribute.ExpectsConsonant);
                modifiedAttrs.add(PhoneticAttribute.ExpectsVowel);
                // TODO: find a better way for this.
                modifiedAttrs.add(PhoneticAttribute.CannotTerminate);
                break;
            case Doubling:
                modifiedSeq.append(alphabet.lastChar(modifiedSeq));
                originalAttrs.add(PhoneticAttribute.ExpectsConsonant);
                modifiedAttrs.add(PhoneticAttribute.ExpectsVowel);
                modifiedAttrs.add(PhoneticAttribute.CannotTerminate);
                break;
            case LastVowelDrop:
                TurkicLetter lastLetter = alphabet.getLastLetter(modifiedSeq);
                if (lastLetter.isVowel()) {
                    modifiedSeq.deleteCharAt(modifiedSeq.length() - 1);
                    modifiedAttrs.add(PhoneticAttribute.ExpectsConsonant);
                    modifiedAttrs.add(PhoneticAttribute.CannotTerminate);
                } else {
                    modifiedSeq.deleteCharAt(modifiedSeq.length() - 2);
                    if (!dicItem.primaryPos.equals(PrimaryPos.Verb)) {
                        originalAttrs.add(PhoneticAttribute.ExpectsConsonant);
                    } else {
                        unmodifiedRootState = morphotactics.verbLastVowelDropUnmodRoot_S;
                        modifiedRootState = morphotactics.verbLastVowelDropModRoot_S;
                    }
                    modifiedAttrs.add(PhoneticAttribute.ExpectsVowel);
                    modifiedAttrs.add(PhoneticAttribute.CannotTerminate);
                }
                break;
            case InverseHarmony:
                originalAttrs.add(PhoneticAttribute.LastVowelFrontal);
                originalAttrs.remove(PhoneticAttribute.LastVowelBack);
                modifiedAttrs.add(PhoneticAttribute.LastVowelFrontal);
                modifiedAttrs.remove(PhoneticAttribute.LastVowelBack);
                break;
            case ProgressiveVowelDrop:
                if (modifiedSeq.length() > 1) {
                    modifiedSeq.deleteCharAt(modifiedSeq.length() - 1);
                    if (alphabet.containsVowel(modifiedSeq)) {
                        modifiedAttrs = calculateAttributes(modifiedSeq);
                    }
                    modifiedAttrs.add(PhoneticAttribute.LastLetterDropped);
                }
                break;
            default:
                break;
        }
    }
    if (unmodifiedRootState == null) {
        unmodifiedRootState = morphotactics.getRootState(dicItem, originalAttrs);
    }
    StemTransition original = new StemTransition(dicItem.root, dicItem, originalAttrs, unmodifiedRootState);
    // if modified root state is not defined in the switch block, get it from morphotactics.
    if (modifiedRootState == null) {
        modifiedRootState = morphotactics.getRootState(dicItem, modifiedAttrs);
    }
    StemTransition modified = new StemTransition(modifiedSeq.toString(), dicItem, modifiedAttrs, modifiedRootState);
    if (original.equals(modified)) {
        return Collections.singletonList(original);
    }
    return Lists.newArrayList(original, modified);
}