How to highlight Boolean FuzzyQueries in Lucene - boost must be a positive float?

Question

I'm trying to be nice for users that make a lot of typos (like myself).

I try to create a simple search page for some data. I build FuzzyQuery s in a BooleanQuery because I would like the user to make typos, for example this:

    BooleanQuery.Builder builder = new BooleanQuery.Builder();

    builder.add(new FuzzyQuery(new Term("body", "pzza")), BooleanClause.Occur.SHOULD);
    builder.add(new FuzzyQuery(new Term("body", "tcyoon")), BooleanClause.Occur.SHOULD);

    BooleanQuery query = builder.build();

Searching works as expected, but the code I got from the Lucene 8.5 API docs to build the highlighting fails:

    SimpleHTMLFormatter htmlFormatter = new SimpleHTMLFormatter();
    Highlighter highlighter = new Highlighter(htmlFormatter, new QueryScorer(query));
    for (int i = 0; i < hits.length; i++) {
        int id = hits[i].doc;
        Document doc = searcher.doc(id);
        System.out.println("HIT:" +  doc.get("url"));
        String text = doc.get("body");
        TokenStream tokenStream = TokenSources.getAnyTokenStream(searcher.getIndexReader(), id, "body", analyzer);
        TextFragment[] frag = highlighter.getBestTextFragments(tokenStream, text, false, 10);//highlighter.getBestFragments(tokenStream, text, 3, "...");
        for (int j = 0; j < frag.length; j++) {
            if ((frag[j] != null) && (frag[j].getScore() > 0)) {
                System.out.println((frag[j].toString()));
            }
        }
    }

With error:

java.lang.IllegalArgumentException: boost must be a positive float, got -1.0
    at org.apache.lucene.search.BoostQuery.<init>(BoostQuery.java:44)
    at org.apache.lucene.search.ScoringRewrite$1.addClause(ScoringRewrite.java:69)
    at org.apache.lucene.search.ScoringRewrite$1.addClause(ScoringRewrite.java:54)
    at org.apache.lucene.search.ScoringRewrite.rewrite(ScoringRewrite.java:117)
    at org.apache.lucene.search.highlight.WeightedSpanTermExtractor.extract(WeightedSpanTermExtractor.java:246)
    at org.apache.lucene.search.highlight.WeightedSpanTermExtractor.extract(WeightedSpanTermExtractor.java:135)
    at org.apache.lucene.search.highlight.WeightedSpanTermExtractor.getWeightedSpanTerms(WeightedSpanTermExtractor.java:530)
    at org.apache.lucene.search.highlight.QueryScorer.initExtractor(QueryScorer.java:218)
    at org.apache.lucene.search.highlight.QueryScorer.init(QueryScorer.java:186)
    at org.apache.lucene.search.highlight.Highlighter.getBestTextFragments(Highlighter.java:201)

The code uses a deprecated method, but I took it straight from the documentation.

Can somebody explain why I get this error? How can I create a highlighter that works with this query construction? Or do I need a different Query ?

Answer 1

The following approach to highlighting uses Lucene v8.5.0 with the question's fuzzy boolean example.

The results look like this, in my stripped-down demo (but you can refine how the highlighted fragments are displayed, of course):

The highlighting code is as follows:

import java.io.IOException;
import org.apache.lucene.document.Document;
import org.apache.lucene.analysis.TokenStream;
import org.apache.lucene.analysis.Analyzer;
import org.apache.lucene.search.Query;
import org.apache.lucene.search.IndexSearcher;
import org.apache.lucene.search.ScoreDoc;
import org.apache.lucene.search.highlight.SimpleHTMLFormatter;
import org.apache.lucene.search.highlight.Highlighter;
import org.apache.lucene.search.highlight.QueryScorer;
import org.apache.lucene.search.highlight.TokenSources;
import org.apache.lucene.search.highlight.TextFragment;
import org.apache.lucene.search.highlight.InvalidTokenOffsetsException;

public class CustomHighlighter {

    private static final String PRE_TAG = "<span class=\"hilite\">";
    private static final String POST_TAG = "</span>";

    public static String[] highlight(Query query, IndexSearcher searcher,
            Analyzer analyzer, ScoreDoc hit, String fieldName)
            throws IOException, InvalidTokenOffsetsException {
        SimpleHTMLFormatter htmlFormatter = new SimpleHTMLFormatter(PRE_TAG, POST_TAG);
        Highlighter highlighter = new Highlighter(htmlFormatter, new QueryScorer(query));
        int id = hit.doc;
        Document doc = searcher.doc(id);

        String text = doc.get(fieldName);

        TokenStream tokenStream = TokenSources.getTokenStream(fieldName,
                searcher.getIndexReader().getTermVectors(id), text, analyzer, -1);
        int maxNumFragments = 10;

        boolean mergeContiguousFragments = Boolean.TRUE;
        TextFragment[] frags = highlighter.getBestTextFragments(tokenStream,
                text, mergeContiguousFragments, maxNumFragments);

        String[] highlightedText = new String[frags.length];
        for (int i = 0; i < frags.length; i++) {
            highlightedText[i] = frags[i].toString();
        }
        // control how you handle each fragment for display...
        //for (TextFragment frag : frags) {
        //    if ((frag != null) && (frag.getScore() > 0)) {
        //        highlightedText = frag.toString();
        //    }
        //}
        return highlightedText;
    }

}

The class is used as follows (where SearchResults is just one of my classes for collecting results, for later presentation to the user):

for (ScoreDoc hit : hits) {
    String[] highlightedText = CustomHighlighter.highlight(query, searcher,
            analyzer, hit, field);
    String document = searcher.doc(hit.doc).get("path");
    SearchResults.Match match = new SearchResults.Match(document, highlightedText, hit.score);
    results.getMatches().add(match);
}

And the fuzzy query is this:

private static Query useFuzzyBooleanQuery() {
    BooleanQuery.Builder builder = new BooleanQuery.Builder();
    builder.add(new FuzzyQuery(new Term("contents", "pzza")), BooleanClause.Occur.SHOULD);
    builder.add(new FuzzyQuery(new Term("contents", "tcyoon")), BooleanClause.Occur.SHOULD);
    return builder.build();
}

The above code does not give me any deprecation warnings.

I can't explain why you get that specific "boost" error - I have not seen that myself, and I was not able to recreate it. But I did not try too hard, I confess.