search

turbomaze / word2vecjson

This project represents the 300-dimensional word vectors from word2vec as JSON.
117 stars 34 forks source link

[Question] Source of vectors? #1

Closed mreishus closed 7 years ago

mreishus commented 7 years ago

Thanks for this project. It's interesting to see it all done in javascript. I'm just curious, what is the source of the vectors you have included here? Did you generate them yourself of use a truncated version of a famous source like GoogleNews-vectors-negative300.bin.gz?

turbomaze commented 7 years ago

Yup! You're right. I suppose I should make this clearer in the readme, but the vectors come from the first link I shared, the version you mentioned.

On Mon, Mar 6, 2017, 10:37 mreishus notifications@github.com wrote:

Thanks for this project. It's interesting to see it all done in javascript. I'm just curious, what is the source of the vectors you have included here? Did you generate them yourself of use a truncated version of a famous source like GoogleNews-vectors-negative300.bin.gz?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/turbomaze/word2vecjson/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/AEodKgnzwbtcbHN-EswlmoIHG3SnHVdTks5rjCgrgaJpZM4MUTix .

mreishus commented 7 years ago

Ok, thanks!

kostasx commented 6 years ago

Shit! Seems like patriarchy is baked into the vectors' source...

Similar vectors to 'woman': woman 1 man 0.7664011739467016 girl 0.7494646687908358 ... housewife 0.5463817659534669 victim 0.5450068478979083

:(