关于labels及label_id的设计

kyzhouhzau / BERT-NER

Use Google's BERT for named entity recognition （CoNLL-2003 as the dataset）.

MIT License

1.24k stars 335 forks source link

关于labels及label_id的设计 #18

Open wangyi888 opened 5 years ago

wangyi888 commented 5 years ago

你好，

看到你对BERT-NER的尝试很受启发。有个问题需要请教，BERT论文原文中提到“where no prediction is made for X”，那么我们还是否将labels中加入“X”呢，如果不加入“X”，又该怎么去体现“no prediction is made”呢？

kyzhouhzau commented 5 years ago

Hi @wangyi888 , In my current version, I used X as a classification label. This is actually not completely correct. In my new attempt, I am not consider the X part. I will replace the word directly with the first part after WordPiece tokenization. The code is still in debugging.

kugwzk commented 5 years ago

Hi @kyzhouhzau, I don't know very well, I see the dataset and I couldn't find the label X in ground-truth label, so how should I eval the result?

Wanjun0511 commented 5 years ago

@kyzhouhzau 这块现在有进展了吗？