2011/7/4 Jörn Kottmann <[email protected]>: > On 7/4/11 11:55 AM, Olivier Grisel wrote: >> >> Out of curiosity is the existing OpenNLP coref implementation >> MaxEnt-based or is it rule based like the state of the art StanfordNLP >> implementation? >> >> http://nlp.stanford.edu/software/dcoref.shtml > > As far as I know it is maxent based, but I do not know much about it > or ever used it, hope that will change soon.
Might be interesting to build a clever rule-based feature generators to find highly non linear, non local features as in DCoref and then train a perceptron for the statistical fine tuning on top of those. That would require a lot of work though. Better start by (re-)training the existing code base on some fresh data. -- Olivier http://twitter.com/ogrisel - http://github.com/ogrisel
