mhesabi mhesabi - 19 days ago 10
C# Question

lucene.net match if search term has no space

I'm using lucene.net to perform searchs in posts in my c# asp.net application, This is a sample document in my indexes:

var doc = new Document();
var title = new Field("Title", "the album hardwired to self-destruct released", Field.Store.YES, Field.Index.ANALYZED, Field.TermVector.WITH_POSITIONS_OFFSETS);
title.Boost = 5;
doc.Add(title);
var ns_title = new Field("NoSpace_Title", "thealbumhardwiredtoselfdesctructreleased", Field.Store.NO, Field.Index.ANALYZED, Field.TermVector.WITH_POSITIONS_OFFSETS);
ns_title.Boost = 5;
doc.Add(ns_title);
doc.Add(new Field("Body", "the body text of the post", Field.Store.YES, Field.Index.ANALYZED, Field.TermVector.WITH_POSITIONS_OFFSETS));
doc.Add(new Field("Id", "1", Field.Store.YES, Field.Index.NOT_ANALYZED));
writer.AddDocument(doc);


Problem:

if I search for
self
or
destruct
or
self destruct
I get hit.

if I search for
selfdestruct
I dont get a hit.

The search method:

var searchWords = s.Split(' ').ToList();
var directory = GetDirectory();
var reader = IndexReader.Open(directory, true);
var searcher = new IndexSearcher(reader);
var analyzer = new StandardAnalyzer(Lucene.Net.Util.Version.LUCENE_30);
var parser = new MultiFieldQueryParser(Lucene.Net.Util.Version.LUCENE_30, "Title,NoSpace_Title,Body".Split(','), analyzer);
var booleanQuery = new BooleanQuery();

// Title:selfdestruct*NoSpace_Title:selfdestruct*Body:selfdestruct*
s = string.Join(" ", searchWords.Select(x => x.Contains("*") ? x : x + "*"));
Query query = parser.Parse(QueryParser.Escape(s));
query.Boost = 5;
booleanQuery.Add(query, Occur.SHOULD);

// Title:*selfdestruct*,NoSpace_Title:*selfdestruct*,Body:*selfdestruct*
// (I suppose this should work and get hit but it doesn't)
s = "*" + string.Join("", searchWords) + "*";
Query query2 = parser.Parse(QueryParser.Escape(s));
query2.Boost = 3;
booleanQuery.Add(query2, Occur.SHOULD);

// Title:selfdestruct~0.85 (fuzzy search)
s = string.Join(" ", searchWords.Select(x => x.Contains("~") ? x : x + "~0.85"));
Query query3 = parser.Parse(QueryParser.Escape(s));
booleanQuery.Add(query3, Occur.SHOULD);

var collector = TopScoreDocCollector.Create(1000, true);
searcher.Search(booleanQuery, collector);
var hits = collector.TopDocs().ScoreDocs;
var docs = hits.Select(x => searcher.Doc(x.Doc)).ToList();

Answer

You can support this by adding a ShingleFilter into your analyzer.

ShingleFilter will combine adjacent tokens into single tokens to facilitate searching for them without a space. By default, it will output Unigrams as well (that is, it will also maintain the single tokens). So, when you index "self-destruct", it will index the tokens "self", "destruct", and "selfdestruct".

An easy way to do this without creating your own custom analyzer, is to use ShingleAnalyzerWrapper:

var analyzer = new ShingleAnalyzerWrapper(
        new StandardAnalyzer(Lucene.Net.Util.Version.LUCENE_30),
        2);