Otis Gospodnetic wrote:
Matt,
Erik and I have some code for this in Lucene in Action, but David Spencer did this since the book was published:
http://www.lucenebook.com/blog/announcements/more_like_this.html
If you want an informal way of doing it you're right, just feed the words of the source doc to a query. The doc for the code it is at this easy to remember URL:
http://searchmorph.com/pub/jakarta-lucene-sandbox/contributions/similarity/build/docs/api/org/apache/lucene/search/similar/SimilarityQueries.html#formSimilarQuery(java.lang.String,%20org.apache.lucene.analysis.Analyzer,%20java.lang.String,%20java.util.Set)
Follow Otis's link above to my weblog for the code.
The "MoreLikeThis" stuff is similar but more sophisticated.
Also if you want the IR way I think you'd do a "cosine measure". I know carrot2 has the code - this might be it:
http://www.searchmorph.com/pub/carrot2/jd/com/chilang/carrot/filter/cluster/rough/measure/CosineCoefficient.html
Otis
--- Matt Chaput <[EMAIL PROTECTED]> wrote:
Is there a simple, efficient way to compute similarity of documents indexed with Lucene?
My first, naive idea is to use the entire contents of one document as
a query to the second document, and use the score as a similarity measurement. But I think I'm probably way off base with that.
Can any IR pros set me straight? Thanks very much.
Matt
-- Matt Chaput Word Monkey Side Effects Software Inc.
"A goddamned ray of sunshine all the goddamned time" -- Sparkle Hayter
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
