Examples with LMDirichletSimilarity org.apache.lucene.search.similarities.LMDirichletSimilarity used on opensource projects

Example 1 with LMDirichletSimilarity

use of org.apache.lucene.search.similarities.LMDirichletSimilarity in project elasticsearch by elastic.

the class SimilarityTests method testResolveSimilaritiesFromMapping_LMDirichlet.

public void testResolveSimilaritiesFromMapping_LMDirichlet() throws IOException {
    String mapping = XContentFactory.jsonBuilder().startObject().startObject("type").startObject("properties").startObject("field1").field("type", "text").field("similarity", "my_similarity").endObject().endObject().endObject().endObject().string();
    Settings indexSettings = Settings.builder().put("index.similarity.my_similarity.type", "LMDirichlet").put("index.similarity.my_similarity.mu", 3000f).build();
    IndexService indexService = createIndex("foo", indexSettings);
    DocumentMapper documentMapper = indexService.mapperService().documentMapperParser().parse("type", new CompressedXContent(mapping));
    assertThat(documentMapper.mappers().getMapper("field1").fieldType().similarity(), instanceOf(LMDirichletSimilarityProvider.class));
    LMDirichletSimilarity similarity = (LMDirichletSimilarity) documentMapper.mappers().getMapper("field1").fieldType().similarity().get();
    assertThat(similarity.getMu(), equalTo(3000f));
}

Example 2 with LMDirichletSimilarity

use of org.apache.lucene.search.similarities.LMDirichletSimilarity in project lucene-solr by apache.

the class KNearestNeighborClassifierTest method testBasicUsage.

@Test
public void testBasicUsage() throws Exception {
    LeafReader leafReader = null;
    try {
        MockAnalyzer analyzer = new MockAnalyzer(random());
        leafReader = getSampleIndex(analyzer);
        checkCorrectClassification(new KNearestNeighborClassifier(leafReader, null, analyzer, null, 1, 0, 0, categoryFieldName, textFieldName), TECHNOLOGY_INPUT, TECHNOLOGY_RESULT);
        checkCorrectClassification(new KNearestNeighborClassifier(leafReader, new LMDirichletSimilarity(), analyzer, null, 1, 0, 0, categoryFieldName, textFieldName), TECHNOLOGY_INPUT, TECHNOLOGY_RESULT);
        ClassificationResult<BytesRef> resultDS = checkCorrectClassification(new KNearestNeighborClassifier(leafReader, new BM25Similarity(), analyzer, null, 3, 2, 1, categoryFieldName, textFieldName), TECHNOLOGY_INPUT, TECHNOLOGY_RESULT);
        ClassificationResult<BytesRef> resultLMS = checkCorrectClassification(new KNearestNeighborClassifier(leafReader, new LMDirichletSimilarity(), analyzer, null, 3, 2, 1, categoryFieldName, textFieldName), TECHNOLOGY_INPUT, TECHNOLOGY_RESULT);
        assertTrue(resultDS.getScore() != resultLMS.getScore());
    } finally {
        if (leafReader != null) {
            leafReader.close();
        }
    }
}

Example 3 with LMDirichletSimilarity

use of org.apache.lucene.search.similarities.LMDirichletSimilarity in project Anserini by castorini.

the class SearchTweets method main.

public static void main(String[] args) throws Exception {
    long curTime = System.nanoTime();
    SearchArgs searchArgs = new SearchArgs();
    CmdLineParser parser = new CmdLineParser(searchArgs, ParserProperties.defaults().withUsageWidth(90));
    try {
        parser.parseArgument(args);
    } catch (CmdLineException e) {
        System.err.println(e.getMessage());
        parser.printUsage(System.err);
        System.err.println("Example: SearchTweets" + parser.printExample(OptionHandlerFilter.REQUIRED));
        return;
    }
    LOG.info("Reading index at " + searchArgs.index);
    Directory dir;
    if (searchArgs.inmem) {
        LOG.info("Using MMapDirectory with preload");
        dir = new MMapDirectory(Paths.get(searchArgs.index));
        ((MMapDirectory) dir).setPreload(true);
    } else {
        LOG.info("Using default FSDirectory");
        dir = FSDirectory.open(Paths.get(searchArgs.index));
    }
    IndexReader reader = DirectoryReader.open(dir);
    IndexSearcher searcher = new IndexSearcher(reader);
    if (searchArgs.ql) {
        LOG.info("Using QL scoring model");
        searcher.setSimilarity(new LMDirichletSimilarity(searchArgs.mu));
    } else if (searchArgs.bm25) {
        LOG.info("Using BM25 scoring model");
        searcher.setSimilarity(new BM25Similarity(searchArgs.k1, searchArgs.b));
    } else {
        LOG.error("Error: Must specify scoring model!");
        System.exit(-1);
    }
    RerankerCascade cascade = new RerankerCascade();
    if (searchArgs.rm3) {
        cascade.add(new Rm3Reranker(IndexTweets.ANALYZER, StatusField.TEXT.name, "src/main/resources/io/anserini/rerank/rm3/rm3-stoplist.twitter.txt"));
        cascade.add(new RemoveRetweetsTemporalTiebreakReranker());
    } else {
        cascade.add(new RemoveRetweetsTemporalTiebreakReranker());
    }
    if (!searchArgs.model.isEmpty() && searchArgs.extractors != null) {
        LOG.debug(String.format("Ranklib model used, modeled loaded from %s", searchArgs.model));
        cascade.add(new RankLibReranker(searchArgs.model, StatusField.TEXT.name, searchArgs.extractors));
    }
    FeatureExtractors extractorChain = null;
    if (searchArgs.extractors != null) {
        extractorChain = FeatureExtractors.loadExtractor(searchArgs.extractors);
    }
    if (searchArgs.dumpFeatures) {
        PrintStream out = new PrintStream(searchArgs.featureFile);
        Qrels qrels = new Qrels(searchArgs.qrels);
        cascade.add(new TweetsLtrDataGenerator(out, qrels, extractorChain));
    }
    MicroblogTopicSet topics = MicroblogTopicSet.fromFile(new File(searchArgs.topics));
    PrintStream out = new PrintStream(new FileOutputStream(new File(searchArgs.output)));
    LOG.info("Writing output to " + searchArgs.output);
    LOG.info("Initialized complete! (elapsed time = " + (System.nanoTime() - curTime) / 1000000 + "ms)");
    long totalTime = 0;
    int cnt = 0;
    for (MicroblogTopic topic : topics) {
        long curQueryTime = System.nanoTime();
        Query filter = LongPoint.newRangeQuery(StatusField.ID.name, 0L, topic.getQueryTweetTime());
        Query query = AnalyzerUtils.buildBagOfWordsQuery(StatusField.TEXT.name, IndexTweets.ANALYZER, topic.getQuery());
        BooleanQuery.Builder builder = new BooleanQuery.Builder();
        builder.add(filter, BooleanClause.Occur.FILTER);
        builder.add(query, BooleanClause.Occur.MUST);
        Query q = builder.build();
        TopDocs rs = searcher.search(q, searchArgs.hits);
        List<String> queryTokens = AnalyzerUtils.tokenize(IndexTweets.ANALYZER, topic.getQuery());
        RerankerContext context = new RerankerContext(searcher, query, topic.getId(), topic.getQuery(), queryTokens, StatusField.TEXT.name, filter);
        ScoredDocuments docs = cascade.run(ScoredDocuments.fromTopDocs(rs, searcher), context);
        for (int i = 0; i < docs.documents.length; i++) {
            String qid = topic.getId().replaceFirst("^MB0*", "");
            out.println(String.format("%s Q0 %s %d %f %s", qid, docs.documents[i].getField(StatusField.ID.name).numericValue(), (i + 1), docs.scores[i], searchArgs.runtag));
        }
        long qtime = (System.nanoTime() - curQueryTime) / 1000000;
        LOG.info("Query " + topic.getId() + " (elapsed time = " + qtime + "ms)");
        totalTime += qtime;
        cnt++;
    }
    LOG.info("All queries completed!");
    LOG.info("Total elapsed time = " + totalTime + "ms");
    LOG.info("Average query latency = " + (totalTime / cnt) + "ms");
    reader.close();
    out.close();
}

Example 4 with LMDirichletSimilarity

use of org.apache.lucene.search.similarities.LMDirichletSimilarity in project Anserini by castorini.

the class SearchWebCollection method main.

public static void main(String[] args) throws Exception {
    SearchArgs searchArgs = new SearchArgs();
    CmdLineParser parser = new CmdLineParser(searchArgs, ParserProperties.defaults().withUsageWidth(90));
    try {
        parser.parseArgument(args);
    } catch (CmdLineException e) {
        System.err.println(e.getMessage());
        parser.printUsage(System.err);
        System.err.println("Example: SearchWebCollection" + parser.printExample(OptionHandlerFilter.REQUIRED));
        return;
    }
    LOG.info("Reading index at " + searchArgs.index);
    Directory dir;
    if (searchArgs.inmem) {
        LOG.info("Using MMapDirectory with preload");
        dir = new MMapDirectory(Paths.get(searchArgs.index));
        ((MMapDirectory) dir).setPreload(true);
    } else {
        LOG.info("Using default FSDirectory");
        dir = FSDirectory.open(Paths.get(searchArgs.index));
    }
    Similarity similarity = null;
    if (searchArgs.ql) {
        LOG.info("Using QL scoring model");
        similarity = new LMDirichletSimilarity(searchArgs.mu);
    } else if (searchArgs.bm25) {
        LOG.info("Using BM25 scoring model");
        similarity = new BM25Similarity(searchArgs.k1, searchArgs.b);
    } else {
        LOG.error("Error: Must specify scoring model!");
        System.exit(-1);
    }
    RerankerCascade cascade = new RerankerCascade();
    boolean useQueryParser = false;
    if (searchArgs.rm3) {
        cascade.add(new Rm3Reranker(new EnglishAnalyzer(), FIELD_BODY, "src/main/resources/io/anserini/rerank/rm3/rm3-stoplist.gov2.txt"));
        useQueryParser = true;
    } else {
        cascade.add(new IdentityReranker());
    }
    FeatureExtractors extractors = null;
    if (searchArgs.extractors != null) {
        extractors = FeatureExtractors.loadExtractor(searchArgs.extractors);
    }
    if (searchArgs.dumpFeatures) {
        PrintStream out = new PrintStream(searchArgs.featureFile);
        Qrels qrels = new Qrels(searchArgs.qrels);
        cascade.add(new WebCollectionLtrDataGenerator(out, qrels, extractors));
    }
    Path topicsFile = Paths.get(searchArgs.topics);
    if (!Files.exists(topicsFile) || !Files.isRegularFile(topicsFile) || !Files.isReadable(topicsFile)) {
        throw new IllegalArgumentException("Topics file : " + topicsFile + " does not exist or is not a (readable) file.");
    }
    TopicReader tr = (TopicReader) Class.forName("io.anserini.search.query." + searchArgs.topicReader + "TopicReader").getConstructor(Path.class).newInstance(topicsFile);
    SortedMap<Integer, String> topics = tr.read();
    final long start = System.nanoTime();
    SearchWebCollection searcher = new SearchWebCollection(searchArgs.index);
    searcher.search(topics, searchArgs.output, similarity, searchArgs.hits, cascade, useQueryParser, searchArgs.keepstop);
    searcher.close();
    final long durationMillis = TimeUnit.MILLISECONDS.convert(System.nanoTime() - start, TimeUnit.NANOSECONDS);
    LOG.info("Total " + topics.size() + " topics searched in " + DurationFormatUtils.formatDuration(durationMillis, "HH:mm:ss"));
}

Example 5 with LMDirichletSimilarity

use of org.apache.lucene.search.similarities.LMDirichletSimilarity in project lucene-solr by apache.

the class TestLMDirichletSimilarityFactory method testParameters.

/** dirichlet with parameters */
public void testParameters() throws Exception {
    Similarity sim = getSimilarity("text_params");
    assertEquals(LMDirichletSimilarity.class, sim.getClass());
    LMDirichletSimilarity lm = (LMDirichletSimilarity) sim;
    assertEquals(1000f, lm.getMu(), 0.01f);
}