Details
-
Bug
-
Resolution: Fixed
-
Major
-
6.5.0
-
Untriaged
-
No
Description
Background
FTS should provide updated stemmer (snowball based) libraries to enable the customer for an NLP application.
Problem
The problem appears to be that the stemmer libraries we are using are based on some older algorithm (porter stemmer), while providing the algorithmic results are not the 'naturally' expected results. The example as used by us and the customer is to do a stemming search for the word 'enjoy' and they expect to get back results which contain the stem word and it's derivative, however they are not seeing this.
This seems to be fixed in the more recent version of the porter stemmer algorithm(snowball) which delivers more accurate results.