uima-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jens Grivolla <j+...@grivolla.net>
Subject Re: OpenNLP type system problem
Date Thu, 06 Oct 2011 09:46:07 GMT
On 10/05/2011 10:16 PM, Jörn Kottmann wrote:
> On 10/5/11 9:40 PM, Nicolas Hernandez wrote:
>>
>> In order to make the AE works I try to use the
>> opennlp-uima/descriptors/TypeSystem.xml file. Indeed the
>> opennlp.uima.Chunk type contains a "type" attribute. Personnaly I
>> renamed it in "chunkType" ...
>
> We could just change it, the OpenNLP type system is an example only
> anyway since all types in the OpenNLP AEs can be mapped to a user type
> system by descriptor configuration.

We are currently moving from the Julielab wrappers to the "official" 
UIMA integration.  Given that the postag is now a (string) feature of 
the token, instead of an FSArray of PosTag annotations as in the 
Julielab version, this means we need to change our type system. As we're 
using OpenNLP for much of our base (sentence splitting, tokenization, 
POS tagging, and maybe chunking) we obviously looked at the provided 
type system to base our types on.

I think we'll just go ahead using our own types and thus avoiding the 
problem, but it would probably be best to fix the OpenNLP example TS. 
It might be a bit painful though for people who have already started 
using it, and might get problems if there are different definitions for 
the same type (opennlp.uima.Chunk).

Jens


Mime
View raw message