sent2feature at test_data

[tokawa0213]

Hi,
I read your code at NER using CRF.ipynb and I have a question about using sent2features at test_data.
Is it appropriate to use sent2features at test_data ? At sent2features it seems that you are giving test_data a data we should not know (POS tag of previous and continuing word). Rather than giving the gold data of POS tags, I think that the model should use their own prediction result.
Although most of the pos tagging CRF tutorials does it the same way, I want to ask your opinion and whether it is possible to implement it.
Thank you in advance