amplab / keystone

Simplifying robust end-to-end machine learning on Apache Spark.
http://keystone-ml.org/
Apache License 2.0
470 stars 117 forks source link

Named entity recognition #274

Closed ngarneau closed 8 years ago

ngarneau commented 8 years ago

Hey guys, on the same idea as the POS tagger, here is the NER. As @etrain said, it would be better if the Keystone API wouldn't expose Epic's interfaces so I'll work on this as a next step.

Let me know if you see anything here, I built it pretty much the same way I did with the POS Tagger.

Thanks, Nicolas

etrain commented 8 years ago

A couple of minor comments - in general we're trying to preserve type-safety at the level of pipeline nodes, and I think a generic will solve that problem here (though it may not - the epic type system isn't as crazy as breeze's but it's not the simplest one either)

ngarneau commented 8 years ago

I definitely agree with your comments.

etrain commented 8 years ago

LGTM. Thanks!