Question: Have you tried feature-based approach without fine tuning BERT?

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

Question: Have you tried feature-based approach without fine tuning BERT? #58

Open hkvision opened 5 years ago

hkvision commented 5 years ago

As described in BERT paper Section 5.4, it gives some experiment results of not fine tuning BERT parameters but use BERT as a feature extractor followed by 2 Bi-LSTMs. Wondering have you tried this so that I can have some reference?

kyzhouhzau commented 5 years ago

i tried bert service to extract hidden vectors, good results were obtained, but it was a preliminary experiment and it was not finished.

zwd13122889 commented 4 years ago

Excuse me. Where does the label_test.txt come from? Man made or machine generated?

hkvision commented 4 years ago

@zwd13122889 https://soco.intel.com/photos/184811?photoIndex=61 I suppose this is the output file by the program for the predictions.