I now got it running with 1.5.
You need to do the following.
Use our Parser to parse an input article.
Use the TreebankNameFinder to add names to your parsed article (its now
in the trunk, see OPENNLP-407).
I used this command:
java -cp ... TreebankNameFinder -parse ner/date.bin ner/location.bin
ner/money.bin ner/organization.bin ner/percentage.bin ner/person.bin
ner/time.bin
The names of models does matter. When you get the models from the
website rename them as I did.
Now you need to run the TreebankLinker. It just needs the model
directory and you need to set WNSEARCHDIR
to your wordnet directory, e.g like this -DWNSEARCHDIR=wordnet/dict
Now the TreebankLinker is ready to link mentions together.
Let us know if you have issues to get this running.
Hope this helps,
Jörn
On 12/1/11 9:34 AM, Jörn Kottmann wrote:
Exactly, that is the place to look.
I will give it a try as well, maybe we can write
down a few steps to get it working with the current
models.
Jörn
On 12/1/11 9:25 AM, Ben Podgursky wrote:
Hi,
I just started working with opennlp recently so I could be totally
wrong on
this, but I think the main method in TreebankLinker is a decent
example of
how to use coreference resolution.
Ben
On Wed, Nov 30, 2011 at 4:19 AM, SAMIK CHAKRABORTY<sam...@gmail.com>
wrote:
Hi,
Thanks for your reply. I would love to use this. However, I am kind of
stuck for lack of documentations around it.
I am trying to use the code as is and do some basic analysis.
However, the
problem that I am facing (may be a newbie problem), is the entry
point to
the coref.
i have seen that rest of the tools use kind of a common interface to
interact with them (as documented). I am currently going with the
assumption that it will work the same way as others (may be some
changes).
That said, if you could provide me with the entry point (like a class
name), that would be of great help.
I would love to help with this (as much as I can with my limited
knowledge).
On Wed, Nov 30, 2011 at 2:55 PM, Jörn Kottmann<kottm...@gmail.com>
wrote:
On 11/30/11 7:34 AM, SAMIK CHAKRABORTY wrote:
Hi,
i was looking for the coreference resolution using OpenNLP and
UIMA. I
found out that this does not exist in the opennlp-uima code base.
Also this jira issue does not have any pointers to it:
https://issues.apache.org/**jira/browse/OPENNLP-70<
https://issues.apache.org/jira/browse/OPENNLP-70>
I found out that a lot of work has been done in the package
opennlp.tools.coref.
I want to understand a few things:
1. Is the code in opennlp.tools.coref package is ready for a wider
consumption - other than few items like UIMA annotation creation
and CAS
services?
2. If the answer to the first query is "yes". Can I start with a
simple
use
case like use the package to work on?
3. I didn't find any good entry point for the coref, like a JUnit
test
case
to start with. If anyone can tell me an entry point and some
pointers to
start that would of great help.
We currently cannot train the coref component. For this we would need
to have access to coref training data. Which I do not have.
That is the reason why it cannot be really maintained by the
community.
I have never really used it, but I would like to change that and also
work
on
the training part. I as well need the UIMA Integration.
I assume the code is easy to get stable for production use, but it
might
need
a bit time.
Anyway it would be really great if you could help us to improve the
situation.
Jörn
