Open bratao opened 7 years ago
@bratao Hi, thank you for being interested about this project.
I had compared my project to both CRF++ and python-crfsuite on SIGHAN-05 CWS dataset. And the result shows that they have similar performance on PKU dataset.
But the pure linear-chain-CRF(crf_tagger.py
) is not my ultimate goal. My goal is to write an enhanced CRF that can take real-valued vectors, like word embedding of current token, as feature input.
I know python-crfsuite can also do this, but it cannot fine tune the word embedings during training process. So I started this project.
And I think the main use case for linear_chain_crf could be any sequence-labelling tasks, like Chinese word segmentation, NER or POS-tagging.
Thanks:smile:
Hello,
Thank you for providing this awesome tool.
Do you have any idea how your tools compares to the performance of tools like CRF++ or CRFSuite ?
What is the main use case for linear_chain_crf ??
Thanks