Maxsum returns wrong similarlities

[kunihik0]

Hi, thank you for good KeyBert!
I think the method of getting the distance by index in _maxsum.py is wrong . The problem is that the similarity between the sentence and the candidate is specified with a different index than the actual one, so the wrong similarity is returned.
I think the following points need to be changed.

return [(words_vals[idx], round(float(distances[0][idx]), 4)) for idx in candidate]
to
return [(words_vals[idx], round(float(distances[0][words_idx[idx]]), 4)) for idx in candidate]

https://github.com/MaartenGr/KeyBERT/blob/master/keybert/_maxsum.py#:~:text=return%20%5B(words_vals%5Bidx%5D%2C%20round(float(distances%5B0%5D%5Bidx%5D)%2C%204))%20for%20idx%20in%20candidate%5D

[MaartenGr]

Sorry for the late response and thank you for tracking this down! I'll make sure it gets fixed in the next release.

[kunihik0]

Thank you for your reply.
I had never heard of fork before but I have learned a bit about it.
So I would like you to use this pull request if possible after you have checked.

[MaartenGr]

Thank you for the pull request! It was merged and will be added in the next pypi release.

[kunihik0]

Thanks for checking and merging!