You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@opennlp.apache.org by "Jim - FooBar();" <ji...@gmail.com> on 2012/04/04 16:10:56 UTC
Merging results from several name-finders...
On 04/04/12 14:57, Jörn Kottmann wrote:
> Please make a patch and attach it to some jira issue. You also need to
> set the flag that you
> agree that the code will be licensed under AL 2.0. We don't accept
> code contributions which
> are posted to the mailing list.
>
> More information on how to contribute can be found here:
> http://opennlp.apache.org/get-involved.html
Ok thanks i will...
> I had a quick look at your class. Wouldn't it be better write a
> TokenNameFinder which
> acts like a proxy to a set of name finders and aggregates/aligns the
> result of them?
>
> You can then pass an instance of this new name finder to the existing
> evaluator.
Yes, i can also do that but that sounds even more like a hack! At least
by containing the issue inside TokenNameFinderEvaluator we're saying:
*"Look this is an evaluation issue and has little to do with the
name-finders"*
otherwise we're saying
*"Hmmm, there is a different, better name-finder and all it does that
calls the regular ones!"*
That is slightly misleading isn't it?
The way i set it up we can easily write some code inside the existing
TokenNameFinder class adding some "if"s to handle the case where the
user passes in some more name-finders. IT doesn't even have to break
backwards compatibility...Unless you insist i follow your suggestion, I
will follow my rationale but improve on the code so it properly handles
all cases inside the existing TokenNameFinder class and open a jira to
submit it...
Jim
Re: Merging results from several name-finders...
Posted by Jörn Kottmann <ko...@gmail.com>.
On 04/04/2012 04:10 PM, Jim - FooBar(); wrote:
>> I had a quick look at your class. Wouldn't it be better write a
>> TokenNameFinder which
>> acts like a proxy to a set of name finders and aggregates/aligns the
>> result of them?
>>
>> You can then pass an instance of this new name finder to the existing
>> evaluator.
>
> Yes, i can also do that but that sounds even more like a hack!
Why do you think its a hack? Everyone who runs multiple name finders needs
to merge the results somehow in the end. Its a very common problem
we have not yet solved in OpenNLP.
And most people detect multiple entity types, e.g person and
organization names.
Jörn