Why the predict result is so ridiculous?

AndriyMulyar / semantic-text-similarity

an easy-to-use interface to fine-tuned BERT models for computing semantic similarity in clinical and web text. that's it.

MIT License

215 stars 51 forks source link

Open power76 opened 4 years ago

power76 commented 4 years ago

I have 2 sentences as below: The videocall is unavailable. The videocall is available.

The predict result is 4.71. What's wrong?

weiyx16 commented 3 years ago

I think is normal for current deep learning-based method, for it doesn't truly understand the semantic, but does some statistic only......