Open Access Reference corpus

[jmccrae]

The current guidelines for new synsets, state that the lemma must have at least 100 occurrences in Sketch Engines's TenTen corpus.

https://github.com/globalwordnet/english-wordnet/blob/master/NEW_SYNSETS.md

This corpus is only accessible to paying Sketch Engine customers and so does not really fit with our open-source goals. We should update this to an open access corpus such as the American National Corpus.

Any suggestions?

[arademaker]

1. EWT and Ontonotes from https://github.com/propbank/propbank-release
2. English UD corpora

[jmccrae]

Thanks @arademaker, both of those corpora are quite small and I don't think they would suit our needs.

@fcbond has suggested the use of the CoCA corpus and I think this seems quite suitable.