English NER data about nltk

[kmike]

+1 for CoNLL 2003. I think that training data is more important than pre-training models because you learn by training your own models, not by using some prebuilt ones.

[jnothman]

I think this is a good idea. It might also be interesting to add the entity disambiguation layer annotated by the YAGO-AIDA project (Hoffart et al).

[tgalery]

Sorry to disagree I think having the data might be interesting for a variety of tasks. Also, topic extraction is moving from text classification to entity linking, so datasets like the ones documented in the first task of http://www.wise2013.org/wise2013challenge.html and https://code.google.com/p/wiki-link/wiki/ExpandedDataset might be relevant. Providing an easy way to package and this type of corpora in nltk might be very useful.

[roopalgarg]

+1 for CoNLL 2003. Could really use that for NER work.

[stevenbird]

Closing, as we don't have capacity to take this on.