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

Example 1 with StemAndEnding

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

the class UnidentifiedTokenAnalyzer method tryNumeral.

private List<SingleAnalysis> tryNumeral(Token token) {
    String s = token.getText();
    s = s.toLowerCase(TurkishAlphabet.TR);
    StemAndEnding se = getFromNumeral(s);
    String lemma;
    if (se.stem.endsWith(".")) {
        String ss = se.stem.substring(0, se.stem.length() - 1);
        lemma = numeralEndingMachine.find(ss);
        lemma = ordinalMap.get(lemma);
    } else {
        lemma = numeralEndingMachine.find(se.stem);
    }
    List<SingleAnalysis> results = Lists.newArrayListWithCapacity(1);
    for (Numerals numerals : Numerals.values()) {
        Matcher m = numerals.pattern.matcher(se.stem);
        if (m.find()) {
            String toParse;
            if (se.ending.length() > 0 && lemma.equals("dört") && ALPHABET.isVowel(se.ending.charAt(0))) {
                toParse = "dörd" + se.ending;
            } else {
                toParse = lemma + se.ending;
            }
            List<SingleAnalysis> res = analyzer.analyze(toParse);
            for (SingleAnalysis re : res) {
                if (re.getDictionaryItem().primaryPos != PrimaryPos.Numeral) {
                    continue;
                }
                DictionaryItem runTimeItem = new DictionaryItem(se.stem, se.stem, s + lemma, PrimaryPos.Numeral, numerals.secondaryPos);
                runTimeItem.attributes.add(RootAttribute.Runtime);
                results.add(re.copyFor(runTimeItem, se.stem));
            }
        }
    }
    return results;
}

Example 2 with StemAndEnding

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

the class UnidentifiedTokenAnalyzer method tryWordWithApostrophe.

private List<SingleAnalysis> tryWordWithApostrophe(String word, SecondaryPos secondaryPos) {
    String normalized = TurkishAlphabet.INSTANCE.normalizeApostrophe(word);
    int index = normalized.indexOf('\'');
    if (index <= 0 || index == normalized.length() - 1) {
        return Collections.emptyList();
    }
    String stem = normalized.substring(0, index);
    String ending = normalized.substring(index + 1);
    StemAndEnding se = new StemAndEnding(stem, ending);
    // TODO: should we remove dots with normalization?
    String stemNormalized = TurkishAlphabet.INSTANCE.normalize(se.stem).replaceAll("[.]", "");
    String endingNormalized = TurkishAlphabet.INSTANCE.normalize(se.ending);
    String pronunciation = guessPronunciation(stemNormalized);
    boolean capitalize = secondaryPos == SecondaryPos.ProperNoun || secondaryPos == SecondaryPos.Abbreviation;
    boolean pronunciationPossible = alphabet.containsVowel(pronunciation);
    DictionaryItem item = new DictionaryItem(capitalize ? Turkish.capitalize(normalized) : (pronunciationPossible ? stem : word), stemNormalized, pronunciation, PrimaryPos.Noun, secondaryPos);
    if (!pronunciationPossible) {
        List<SingleAnalysis> result = new ArrayList<>(1);
        result.add(SingleAnalysis.dummy(word, item));
        return result;
    }
    boolean itemDoesNotExist = !lexicon.containsItem(item);
    if (itemDoesNotExist) {
        item.attributes.add(RootAttribute.Runtime);
        analyzer.getStemTransitions().addDictionaryItem(item);
    }
    String toParse = stemNormalized + endingNormalized;
    List<SingleAnalysis> noQuotesParses = analyzer.analyze(toParse);
    if (itemDoesNotExist) {
        analyzer.getStemTransitions().removeDictionaryItem(item);
    }
    List<SingleAnalysis> analyses = noQuotesParses.stream().filter(noQuotesParse -> noQuotesParse.getStem().equals(stemNormalized)).collect(Collectors.toList());
    return analyses;
}

Example 3 with StemAndEnding

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

the class TurkishMorphology method analyzeWordsWithApostrophe.

public List<SingleAnalysis> analyzeWordsWithApostrophe(String word) {
    int index = word.indexOf('\'');
    if (index <= 0 || index == word.length() - 1) {
        return Collections.emptyList();
    }
    StemAndEnding se = new StemAndEnding(word.substring(0, index), word.substring(index + 1));
    String stem = TurkishAlphabet.INSTANCE.normalize(se.stem);
    String withoutQuote = word.replace("'", "");
    List<SingleAnalysis> noQuotesParses = analyzer.analyze(withoutQuote);
    if (noQuotesParses.size() == 0) {
        return Collections.emptyList();
    }
    // words like "Hastanesi'ne". Should we accept Hastanesi or Hastane?
    return noQuotesParses.stream().filter(a -> a.getDictionaryItem().primaryPos == PrimaryPos.Noun && (a.containsMorpheme(TurkishMorphotactics.p3sg) || a.getStem().equals(stem))).collect(Collectors.toList());
}

Example 4 with StemAndEnding

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

the class UnidentifiedTokenAnalyzer method getForRomanNumeral.

private List<SingleAnalysis> getForRomanNumeral(Token token) {
    String content = token.getText();
    StemAndEnding se;
    if (content.contains("'")) {
        int i = content.indexOf('\'');
        se = new StemAndEnding(content.substring(0, i), content.substring(i + 1));
    } else {
        se = new StemAndEnding(content, "");
    }
    String ss = se.stem;
    if (se.stem.endsWith(".")) {
        ss = se.stem.substring(0, se.stem.length() - 1);
    }
    int decimal = TurkishNumbers.romanToDecimal(ss);
    if (decimal == -1) {
        return new ArrayList<>(0);
    }
    String lemma;
    if (se.stem.endsWith(".")) {
        lemma = numeralEndingMachine.find(String.valueOf(decimal));
        lemma = ordinalMap.get(lemma);
    } else {
        lemma = numeralEndingMachine.find(String.valueOf(decimal));
    }
    List<SingleAnalysis> results = Lists.newArrayListWithCapacity(1);
    String toParse;
    if (se.ending.length() > 0 && lemma.equals("dört") && ALPHABET.isVowel(se.ending.charAt(0))) {
        toParse = "dörd" + se.ending;
    } else {
        toParse = lemma + se.ending;
    }
    List<SingleAnalysis> res = analyzer.analyze(toParse);
    for (SingleAnalysis re : res) {
        if (re.getDictionaryItem().primaryPos != PrimaryPos.Numeral) {
            continue;
        }
        DictionaryItem runTimeItem = new DictionaryItem(se.stem, se.stem, content + lemma, PrimaryPos.Numeral, SecondaryPos.RomanNumeral);
        runTimeItem.attributes.add(RootAttribute.Runtime);
        results.add(re.copyFor(runTimeItem, se.stem));
    }
    return results;
}