Hi all,
we never really took care for the coref component here.
There reason I never worked on it is that I do not have any training
data to get it re-trained. The training is also not trivial, since it
requires
wordfreak and maybe some code I do not even have. We also didn't do
any testing.
I just saw that CONLL 2011 was about coreference resolution and they
actually offer training data which can at least be acquired without paying
for it.
Should we rescue the coref component with this training data?
Jörn
