epfml / sent2vec

General purpose unsupervised sentence representations
Other
1.19k stars 256 forks source link

nnsent returns similarity above 1 #94

Closed tqx94 closed 4 years ago

tqx94 commented 4 years ago

Hi, I ran the nnsent commands, and there were instances where I got results above 1.

Does that means the embeddings for sent2vec are not normalised?

Thanks, Farren

mpagli commented 4 years ago

Indeed they are not normalized.

tqx94 commented 4 years ago

Hi, thanks for getting back! do you have suggestions on how I could normalise the cosine similarity? I’m not very familiar with this.