lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Allen Atamer" <aata...@casebank.com>
Subject help with DictionaryFilter replacing an acronym token with multiple tokens
Date Thu, 24 Sep 2009 20:06:06 GMT
Hello List members,

 

Please help me to fix a problem in my DictionaryFilter class.  It is
used to map acronyms, abbreviations, synonyms, etc. to one common root
word/phrase for easy searching.  For example, "temp" is an abbreviation
for "temperature".  One-to-one substitutions work without problem in
this code.  However one-to-many substitution isn't working properly.
For example, the search term "lop" is converted to "low oil pressure".
However, in the database, instances of "lop" are not being converted.
I need Lucene search to pick up both.  Right now my search term "lop"
only picks up documents with "low oil pressure" in them.

 

The log trace of the indexer is as follows:

2009-09-24 15:40:32,774 [     main] DEBUG DictionaryFilter  -  NEWTOKEN
[low] , start:17, end:19

2009-09-24 15:40:32,774 [     main] DEBUG DictionaryFilter  -  NEWTOKEN
[oil] , start:20, end:22

2009-09-24 15:40:32,774 [     main] DEBUG DictionaryFilter  -  NEWTOKEN
[pressure] , start:23, end:30

2009-09-24 15:40:32,774 [     main] INFO  AnalyzerUtils  - [fault]
[mnemonics] [lop] [sdn]

 

So you can see the DictionaryFilter makes the substitutions, but the
Analyzer doesn't keep the new tokens. When I scan through the terms with
AnalyzerUtils, the original token [lop] is still there!  

 

Please help.  Thank you.

 

Below is the logic for DictionaryFilter.next():

 

    public Token next()

        throws java.io.IOException {

        if (!tokenQueue.isEmpty()) {

            return (Token) tokenQueue.pop();

        }

 

        Token reusableToken = new Token();

        Token token = input.next(reusableToken);

 

        if ((dictionary == null) || (token == null) || !processField) {

            return token;

        }

 

        TermData t = (TermData) dictionary.get(token.term());

 

        if (t != null && t.getTeach() != null) {

                  Token result = processTeachToken(t.getTeach(), token);

                  token = null;

                  return result;

            } else if (t != null) { 

                  // return original because there's nothing better to
go on.

                  return token;

            }

            ... // other logic related to spell-checking

      }

 

And here's the processTeachToken() function, which is where the
substitution happens:

 

    private Token processTeachToken(String teachString, Token original)
{

        StringTokenizer tokenizer = new StringTokenizer(teachString, "
");

 

        int start = original.startOffset();

        int positionIncrement = original.getPositionIncrement();

 

        while (tokenizer.hasMoreTokens()) {

            String partToken = tokenizer.nextToken();

 

            if (partToken.equals("")) {

                  throw new RuntimeException("TextClassifier failed");

            }

 

//          Token newToken = new Token(partToken, start,

//          (start + partToken.length()) - 1);

                  Token newToken = new Token(start, (start +
partToken.length()) - 1);

//                newToken.setTermLength(partToken.length());

                  LuceneUtils.copyTermBuffer(newToken, partToken);

            log.debug(" NEWTOKEN [" + partToken + "] , start:"

                + newToken.startOffset() + ", end:" +
newToken.endOffset());

            newToken.setPositionIncrement(positionIncrement);

            tokenQueue.push(newToken);

            start += partToken.length();

            positionIncrement++;

        }

 

        Token result = (Token) tokenQueue.pop();

 

        return result;

    }

 

LuceneUtils.copyTermBuffer:

 

      public static void copyTermBuffer(Token term, String copyTerm) {

        if (term.termBuffer().length < copyTerm.length()) {

                  term.resizeTermBuffer(copyTerm.length());

            }

            

        term.setTermLength(copyTerm.length());

 

            char [] termBuffer = term.termBuffer();

            for (int i = 0; i < copyTerm.length(); i++) {

                  termBuffer[i] = copyTerm.charAt(i);

            }

      }


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message