Foros de discusión

Please upgrade Lucene to Support BM25

thumbnail
Carlos Santos, modificado hace 8 años.

Please upgrade Lucene to Support BM25

New Member Mensajes: 15 Fecha de incorporación: 18/09/15 Mensajes recientes
Liferay 6.2 uses Lucene 3.5.0 that has more than 4 years old (see http://archive.apache.org/dist/lucene/java/).
From Lucene 4.0 it was added support to sophisticated ranking functions as BM25 and contains many other benefits.
The upgrade would allow many necessary improvements in the ranking system of Liferay search.

From our tests (in Liferay CE 6.2) for example the search for "Lord of the Rings" (without commas) retrieve documents that contain "of" or "the" as part of words as in "Kristoff" or "synthetic".
Other than that, documents with all 4 words in the title are not ranked first than documents with "of" as in "Kristoff" in the description of the document.
Many issues also happen in search for phrases using commas.
These are only some examples of issues we found with the ranking used in the search.

All these ranking problems could be solved with the use of BM25.
I'm sure such benefit would be very welcome to other developers and users of Liferay as well.
Thanks for your attention!
thumbnail
David H Nebinger, modificado hace 8 años.

RE: Please upgrade Lucene to Support BM25

Liferay Legend Mensajes: 14918 Fecha de incorporación: 2/09/06 Mensajes recientes
Lucene is not the only game in town. Have you tried SOLR?
thumbnail
Carlos Santos, modificado hace 8 años.

RE: Please upgrade Lucene to Support BM25

New Member Mensajes: 15 Fecha de incorporación: 18/09/15 Mensajes recientes
Hi David!
I haven't yet! This seems a good idea!
I found some links to use SOLR 4+ that comes with Lucene 4+ here:

http://ntrpigo.blogspot.it/2014/08/integrating-solr-43-into-liferay-6.html
http://www.liferay.com/community/wiki/-/wiki/Main/Integrate+Solr+with+Liferay+portal

I'll need to study to see if it works for me but it seems a very good alternative!
Thanks!