kentonl / e2e-coref

End-to-end Neural Coreference Resolution
Apache License 2.0
518 stars 174 forks source link

1-dimensional CNN embeddigs. #95

Open linguist89 opened 3 years ago

linguist89 commented 3 years ago

I'm a little confused as to how the embeddings are structured for each word. I understand that each words has 300D GloVe embeddings and 50D Turian embeddings, but does each word also have the 1D CNN embeddings or are those embeddings only applied to words that are out of vocabulary i.e. if GloVe and Turian embeddings are unavailable?