Examples with MicroblogTopicSet io.anserini.search.MicroblogTopicSet used on opensource projects

Example 1 with MicroblogTopicSet

use of io.anserini.search.MicroblogTopicSet in project Anserini by castorini.

the class DumpTweetsLtrData method main.

public static void main(String[] argv) throws Exception {
    long curTime = System.nanoTime();
    LtrArgs args = new LtrArgs();
    CmdLineParser parser = new CmdLineParser(args, ParserProperties.defaults().withUsageWidth(90));
    try {
        parser.parseArgument(argv);
    } catch (CmdLineException e) {
        System.err.println(e.getMessage());
        parser.printUsage(System.err);
        System.err.println("Example: DumpTweetsLtrData" + parser.printExample(OptionHandlerFilter.REQUIRED));
        return;
    }
    LOG.info("Reading index at " + args.index);
    Directory dir = FSDirectory.open(Paths.get(args.index));
    IndexReader reader = DirectoryReader.open(dir);
    IndexSearcher searcher = new IndexSearcher(reader);
    if (args.ql) {
        LOG.info("Using QL scoring model");
        searcher.setSimilarity(new LMDirichletSimilarity(args.mu));
    } else if (args.bm25) {
        LOG.info("Using BM25 scoring model");
        searcher.setSimilarity(new BM25Similarity(args.k1, args.b));
    } else {
        LOG.error("Error: Must specify scoring model!");
        System.exit(-1);
    }
    Qrels qrels = new Qrels(args.qrels);
    FeatureExtractors extractors = null;
    if (args.extractors != null) {
        extractors = FeatureExtractors.loadExtractor(args.extractors);
    }
    PrintStream out = new PrintStream(new FileOutputStream(new File(args.output)));
    RerankerCascade cascade = new RerankerCascade();
    cascade.add(new RemoveRetweetsTemporalTiebreakReranker());
    cascade.add(new TweetsLtrDataGenerator(out, qrels, extractors));
    MicroblogTopicSet topics = MicroblogTopicSet.fromFile(new File(args.topics));
    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, args.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);
        cascade.run(ScoredDocuments.fromTopDocs(rs, searcher), context);
        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();
}