5 comments

[ 3.3 ms ] story [ 26.5 ms ] thread
A while ago, I tested Stanford's Core NLP Open IE implementation and UW (now AllenAII)'s version[1]. The UW version was much better.

The paper for this make it look pretty good compared to that. I'll be trying it.

[1] https://github.com/allenai/openie-standalone

I came to the same conclusion. OTOH OpenIE is around 4x slower than CoreNLP and the memory requirements can also be a problem (~4GB JVM heap space).

I would be really interested on how ClausIE stacks up to that.

ClausIE also seems to be using CoreNLP _and_ OpenNLP.