uima-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Peter Kl├╝gl <peter.klu...@averbis.com>
Subject Re: Question
Date Wed, 07 Oct 2015 17:44:21 GMT

negated match conditions are not supported in UIMA Ruta and there are no 
plans yet to support them in near future for different reasons. If 
supported in your example, the negation would match on all possible 
combinations of all possible annotations within the sentence that do not 
fulfill the sequential pattern. The problem with true negatives in 
information extraction...

You need to move the negation to the condition part. The easiest way to 
solve your problem is to use two rules. You could do something like:
(not tested right now)

DECLARE ChunkPattern;
Chunk.chunk=="NP"+ Chunk.chunk=="VP"+ Chunk.chunk == "ADJP"{-> 
MARK(ChunkPattern, 1,3)};
Sentence{-CONTAINS(ChunkPattern) -> CREATE(Error, "name"= "SVO", "kind" 
= "Simple/Direct")};

DECLARE ChunkSent;
Sentence{-> ChunkSent} <-{Chunk.chunk=="NP"+ Chunk.chunk=="VP"+ 
Chunk.chunk == "ADJP";};
Sentence{-PARTOF(ChunkSent) -> CREATE(Error, "name"= "SVO", "kind" = 

The helper annotation types can also contain a generic feature for 
storing the string values of the created error annotation.

btw, writing the complete long name for referring to a type can be very 
annoying. You maybe want to take a look at IMPORT FROM AS for aliasing 
or at the type variables.

Let me know if this makes sense for you.



Am 07.10.2015 um 00:25 schrieb Juan Ignacio Velez:
> Hi, i'm developing grammar rules with UIMA Ruta and i have a question. I
> can't write a NOT clause with "-" character. The rule is this:
> edu.isistan.uima.unified.typesystems.nlp.Sentence{ -> CREATE(Error, "name"
> = "SVO", "kind" = "Simple/Direct")} <- {
> - (
>      (edu.isistan.uima.unified.typesystems.nlp.Chunk.chunk == "NP")+
>      (edu.isistan.uima.unified.typesystems.nlp.Chunk.chunk == "VP")+
>      (edu.isistan.uima.unified.typesystems.nlp.Chunk.chunk == "ADJP")
> );
> };
> I appreciate your help.
> Thanks a lot!
> Juan Ignacio

View raw message