Token level coref_clusters access raise exceptions sometimes about neuralcoref

[thomwolf]

Hum maybe that a problem of encoding, do you use python 3 or 2?

[theSage21]

I'm sorry. I should have included that.
Ubuntu 18.04, Python 3.6.2

[theSage21]

Ok. So this happens whenever there's no actual coref in the text.

import spacy
nlp = spacy.load('en_coref_sm')
doc = nlp('john was walking down the road.')
doc[0]._.coref_clusters

Could be because of https://github.com/huggingface/neuralcoref/blob/master/neuralcoref/neuralcoref.pyx#L802

I don't know enough about packaging and cython to verify if this is indeed the problem. My changes are not picked up by cython (?). If it is the issue, I'll be happy to put in a pull request.

[thomwolf]

Oh yes, that's probably the issue. I will fix it tomorrow!

If you want to fix it yourself, you have to recompile the model with NeuralCoref CLI packaging script but I haven't documented that script yet so if you are not familiar with packaging a spaCy model it may be a bit difficult to use right now. Will add doc on this in the next release.

[thomwolf]

Hi @theSage21, I finally found time to work on NeuralCoref.
This should be fixed in the new release (4.0).

[rudraksh97]

I am using en_coref_md. I am getting this error. How to resolve this?

[floraliu1011]

I am using en and I also got this problem