Examples with ReverseStringFilter org.apache.lucene.analysis.reverse.ReverseStringFilter used on opensource projects

Example 1 with ReverseStringFilter

use of org.apache.lucene.analysis.reverse.ReverseStringFilter in project elasticsearch by elastic.

the class NoisyChannelSpellCheckerTests method testMultiGenerator.

public void testMultiGenerator() throws IOException {
    RAMDirectory dir = new RAMDirectory();
    Map<String, Analyzer> mapping = new HashMap<>();
    mapping.put("body_ngram", new Analyzer() {

        @Override
        protected TokenStreamComponents createComponents(String fieldName) {
            Tokenizer t = new StandardTokenizer();
            ShingleFilter tf = new ShingleFilter(t, 2, 3);
            tf.setOutputUnigrams(false);
            return new TokenStreamComponents(t, new LowerCaseFilter(tf));
        }
    });
    mapping.put("body", new Analyzer() {

        @Override
        protected TokenStreamComponents createComponents(String fieldName) {
            Tokenizer t = new StandardTokenizer();
            return new TokenStreamComponents(t, new LowerCaseFilter(t));
        }
    });
    mapping.put("body_reverse", new Analyzer() {

        @Override
        protected TokenStreamComponents createComponents(String fieldName) {
            Tokenizer t = new StandardTokenizer();
            return new TokenStreamComponents(t, new ReverseStringFilter(new LowerCaseFilter(t)));
        }
    });
    PerFieldAnalyzerWrapper wrapper = new PerFieldAnalyzerWrapper(new WhitespaceAnalyzer(), mapping);
    IndexWriterConfig conf = new IndexWriterConfig(wrapper);
    IndexWriter writer = new IndexWriter(dir, conf);
    String[] strings = new String[] { "Xorr the God-Jewel", "Grog the God-Crusher", "Xorn", "Walter Newell", "Wanda Maximoff", "Captain America", "American Ace", "Wundarr the Aquarian", "Will o' the Wisp", "Xemnu the Titan", "Fantastic Four", "Quasar", "Quasar II" };
    for (String line : strings) {
        Document doc = new Document();
        doc.add(new Field("body", line, TextField.TYPE_NOT_STORED));
        doc.add(new Field("body_reverse", line, TextField.TYPE_NOT_STORED));
        doc.add(new Field("body_ngram", line, TextField.TYPE_NOT_STORED));
        writer.addDocument(doc);
    }
    DirectoryReader ir = DirectoryReader.open(writer);
    LaplaceScorer wordScorer = new LaplaceScorer(ir, MultiFields.getTerms(ir, "body_ngram"), "body_ngram", 0.95d, new BytesRef(" "), 0.5f);
    NoisyChannelSpellChecker suggester = new NoisyChannelSpellChecker();
    DirectSpellChecker spellchecker = new DirectSpellChecker();
    spellchecker.setMinQueryLength(1);
    DirectCandidateGenerator forward = new DirectCandidateGenerator(spellchecker, "body", SuggestMode.SUGGEST_ALWAYS, ir, 0.95, 10);
    DirectCandidateGenerator reverse = new DirectCandidateGenerator(spellchecker, "body_reverse", SuggestMode.SUGGEST_ALWAYS, ir, 0.95, 10, wrapper, wrapper, MultiFields.getTerms(ir, "body_reverse"));
    CandidateGenerator generator = new MultiCandidateGeneratorWrapper(10, forward, reverse);
    Correction[] corrections = suggester.getCorrections(wrapper, new BytesRef("american cae"), generator, 1, 1, ir, "body", wordScorer, 1, 2).corrections;
    assertThat(corrections.length, equalTo(1));
    assertThat(corrections[0].join(new BytesRef(" ")).utf8ToString(), equalTo("american ace"));
    generator = new MultiCandidateGeneratorWrapper(5, forward, reverse);
    corrections = suggester.getCorrections(wrapper, new BytesRef("american ame"), generator, 1, 1, ir, "body", wordScorer, 1, 2).corrections;
    assertThat(corrections.length, equalTo(1));
    assertThat(corrections[0].join(new BytesRef(" ")).utf8ToString(), equalTo("american ace"));
    corrections = suggester.getCorrections(wrapper, new BytesRef("american cae"), forward, 1, 1, ir, "body", wordScorer, 1, 2).corrections;
    // only use forward with constant prefix
    assertThat(corrections.length, equalTo(0));
    corrections = suggester.getCorrections(wrapper, new BytesRef("america cae"), generator, 2, 1, ir, "body", wordScorer, 1, 2).corrections;
    assertThat(corrections.length, equalTo(1));
    assertThat(corrections[0].join(new BytesRef(" ")).utf8ToString(), equalTo("american ace"));
    corrections = suggester.getCorrections(wrapper, new BytesRef("Zorr the Got-Jewel"), generator, 0.5f, 4, ir, "body", wordScorer, 0, 2).corrections;
    assertThat(corrections.length, equalTo(4));
    assertThat(corrections[0].join(new BytesRef(" ")).utf8ToString(), equalTo("xorr the god jewel"));
    assertThat(corrections[1].join(new BytesRef(" ")).utf8ToString(), equalTo("zorr the god jewel"));
    assertThat(corrections[2].join(new BytesRef(" ")).utf8ToString(), equalTo("four the god jewel"));
    corrections = suggester.getCorrections(wrapper, new BytesRef("Zorr the Got-Jewel"), generator, 0.5f, 1, ir, "body", wordScorer, 1.5f, 2).corrections;
    assertThat(corrections.length, equalTo(1));
    assertThat(corrections[0].join(new BytesRef(" ")).utf8ToString(), equalTo("xorr the god jewel"));
    corrections = suggester.getCorrections(wrapper, new BytesRef("Xor the Got-Jewel"), generator, 0.5f, 1, ir, "body", wordScorer, 1.5f, 2).corrections;
    assertThat(corrections.length, equalTo(1));
    assertThat(corrections[0].join(new BytesRef(" ")).utf8ToString(), equalTo("xorr the god jewel"));
    // Test a special case where one of the suggest term is unchanged by the postFilter, 'II' here is unchanged by the reverse analyzer.
    corrections = suggester.getCorrections(wrapper, new BytesRef("Quazar II"), generator, 1, 1, ir, "body", wordScorer, 1, 2).corrections;
    assertThat(corrections.length, equalTo(1));
    assertThat(corrections[0].join(new BytesRef(" ")).utf8ToString(), equalTo("quasar ii"));
}

Example 2 with ReverseStringFilter

use of org.apache.lucene.analysis.reverse.ReverseStringFilter in project elasticsearch by elastic.

the class EdgeNGramTokenFilterFactory method create.

@Override
public TokenStream create(TokenStream tokenStream) {
    TokenStream result = tokenStream;
    // side=BACK is not supported anymore but applying ReverseStringFilter up-front and after the token filter has the same effect
    if (side == SIDE_BACK) {
        result = new ReverseStringFilter(result);
    }
    result = new EdgeNGramTokenFilter(result, minGram, maxGram);
    // side=BACK is not supported anymore but applying ReverseStringFilter up-front and after the token filter has the same effect
    if (side == SIDE_BACK) {
        result = new ReverseStringFilter(result);
    }
    return result;
}

Example 3 with ReverseStringFilter

use of org.apache.lucene.analysis.reverse.ReverseStringFilter in project jackrabbit-oak by apache.

the class DefaultAnalyzersConfigurationTest method setUp.

@Before
public void setUp() throws Exception {
    this.exactPathAnalyzer = new Analyzer() {

        @Override
        protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
            Tokenizer source = new KeywordTokenizer(reader);
            return new TokenStreamComponents(source);
        }
    };
    this.parentPathIndexingAnalyzer = new Analyzer() {

        @Override
        protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
            Tokenizer source = new KeywordTokenizer(reader);
            return new TokenStreamComponents(source);
        }
    };
    this.parentPathSearchingAnalyzer = new Analyzer() {

        @Override
        protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
            Tokenizer source = new KeywordTokenizer(reader);
            TokenStream filter = new ReverseStringFilter(Version.LUCENE_47, source);
            filter = new PatternReplaceFilter(filter, Pattern.compile("[^\\/]+\\/"), "", false);
            filter = new ReverseStringFilter(Version.LUCENE_47, filter);
            return new TokenStreamComponents(source, filter);
        }
    };
    this.directChildrenPathIndexingAnalyzer = new Analyzer() {

        @Override
        protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
            Tokenizer source = new KeywordTokenizer(reader);
            TokenStream filter = new ReverseStringFilter(Version.LUCENE_47, source);
            filter = new LengthFilter(Version.LUCENE_47, filter, 2, Integer.MAX_VALUE);
            filter = new PatternReplaceFilter(filter, Pattern.compile("([^\\/]+)(\\/)"), "$2", false);
            filter = new PatternReplaceFilter(filter, Pattern.compile("(\\/)(.+)"), "$2", false);
            filter = new ReverseStringFilter(Version.LUCENE_47, filter);
            return new TokenStreamComponents(source, filter);
        }
    };
    this.directChildrenPathSearchingAnalyzer = new Analyzer() {

        @Override
        protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
            Tokenizer source = new KeywordTokenizer(reader);
            return new TokenStreamComponents(source);
        }
    };
    this.allChildrenPathIndexingAnalyzer = new Analyzer() {

        @Override
        protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
            Tokenizer source = new PathHierarchyTokenizer(reader);
            TokenStream filter = new PatternCaptureGroupTokenFilter(source, false, Pattern.compile("((\\/).*)"));
            filter = new RemoveDuplicatesTokenFilter(filter);
            return new TokenStreamComponents(source, filter);
        }
    };
    this.allChildrenPathSearchingAnalyzer = new Analyzer() {

        @Override
        protected TokenStreamComponents createComponents(String fieldName, Reader reader) {
            Tokenizer source = new KeywordTokenizer(reader);
            return new TokenStreamComponents(source);
        }
    };
}