Li Xiaoyu

Agree no more. The performance was severely pulled down especially when tagging massive short texts. It took about 40 minutes to process 10K tweets in Tesla V100 GPU, but only...

I did some more performance tests. I added a Boolean param in the token(), ner(), and pos() functions somewhat like this: ______________________________________________________________ def _posdep_sent(self, in_sent, is_load_weights=True): # assuming input is...