Hi,
I have been testing some rules for three-tokens sequences using
unification. I describe here my case and what I have found.
I want to match a three-tokens sequence: determinant + possessive + noun
(or adjective). There are two features in unification: gender (masc./fem.)
and number (sing./pl.). Every token of the sequence has four possible
forms: masc sing, masc pl, fem sing, fem pl. In total we have 4x4x4=64
combinations for the full sequence. Of these 64 combinations, only 4 are
grammatically correct: those that agree completely.
When I try to match these 4 correct combinations with unification,
everything is OK. I get 4 matches and 60 non-matches. The problem appears
when I try to match the incorrect 60 combinations with <unify
negate="yes">. In this case I get 36 matches (not the expected 60). I also
discovered that the 24 missing combinations (or most of them) cause the
"out of bonds exception" that appeared before we changed the Unifier code.
I don't fully understand the Unifier.java code. The relation between
negated and non-negated unification seems straightforward, but the results
are bizarre. Any help?
Regards,
Jaume Ortolà
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Languagetool-devel mailing list
Languagetool-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/languagetool-devel
