lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Steven Rowe <sar...@syr.edu>
Subject Re: Is there a way to get absolutely exact phrase matching (no stop words, etc)
Date Mon, 24 Oct 2005 20:32:58 GMT
Hi Bob,

StandardAnalyzer filters the token stream created by StandardTokenizer 
through StandardFilter, LowercaseFilter, and then StopFilter.  Unless 
you supply a stoplist to the StandardAnalyzer constructor, you get the 
default set of English stopwords, from StopAnalyzer:

   public static final String[] ENGLISH_STOP_WORDS = {
     "a", "an", "and", "are", "as", "at", "be", "but", "by",
     "for", "if", "in", "into", "is", "it",
     "no", "not", "of", "on", "or", "s", "such",
     "t", "that", "the", "their", "then", "there", "these",
     "they", "this", "to", "was", "will", "with"
   };

One approach to the problem you're seeing is to advance the token 
position in StopFilter with each stopword encountered, so that phrase 
queries like

    "group effect"

will fail to match against

    "...group of ~a- The effect..."

because the positions for tokens "group" and "effect" would not be adjacent.

(My naive reading of StandardTokenizer.jj, the JavaCC grammar used to 
create StandardTokenizer.java, is that "~a-" will generate a single 
token "a", which will then be filtered out by StopFilter.)

A patch implementing this approach was actually applied to 
StopFilter.java in late 2003, but was reverted shortly afterward, 
because this approach conflicts with the QueryParser and PhraseQuery 
implementations.

See Doug Cutting's description of the problem with the position 
increment modification approach here:
<http://mail-archives.apache.org/mod_mbox/lucene-java-dev/200312.mbox/%3c3FCFB3CA.9000103@lucene.com%3e>

See a colored diff of StopFilter.java, just before and after the 
position increment modification patch was reverted, here:
<http://svn.apache.org/viewcvs.cgi/lucene/java/trunk/src/java/org/apache/lucene/analysis/StopFilter.java?rev=150152&r1=150150&r2=150152&diff_format=h>

This modification is simple and straightforward.  You could make the 
same changes to a local copy of StopFilter (call it PosIncrStopFilter), 
then create and use a StandardAnalyzer clone that uses PosIncrStopFilter 
instead of StopFilter.

Good luck,
Steve Rowe

Bob Mason wrote:
> We have a large body of documents that have xml
> and ocr embedded within one of the xml fields.
> 
> Searches such as "group effect"
> 
> are returning hits for docs such as ones that include the following:
> 
>  ...group of ~a- The effect...
> 
> because, I take it, stop words like 'of' and 'the' and punctuation
> are ignored. Is there anything I can do about this other
> than write an alternative to the Standard Analyzer?
> 
> thanks,
> 
> Bob Mason
> UCSF Tobacco Industy Digital Library

---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org


Mime
View raw message