uclnlp / emoji2vec Goto Github PK

hi, @rockt, I did something like below,

e2v = gensim.models.KeyedVectors.load_word2vec_format('emoji2vec.bin', binary=True)
print (e2v['😂'])

e2v = gensim.models.KeyedVectors.load_word2vec_format('emoji2vec.txt')
print (e2v['😂'])

It gives 2 different result. Do you know why?

The README has beisner [at] princeton [dot] edu as the contact email, but that email no longer will receive email (I graduated a few years ago). A better contact email for me is: ben [dot] a [dot] eisner [at] gmail [dot] com.

Top 4 similar emoji with 😭 by using pre-trained emoji2vec.txt:

😿 0.836434900761
😢 0.689464867115
😹 0.677145779133
😂 0.671003580093

I get it, that crying face emoji is kind of more similar with above 4 emojis, but 😿😢 and 😹😂 represent totally two different part of word sentiment. Positive and Negative! How can I group emojis based by sentiment polarity ?

morn    😭
hurt    😭
crying  😭
tears   😭
loudly crying face  😭
sob 😭
sobbing 😭
sad tears   😭
bawling 😭
sad 😭
melancholy  😭
cry 😭
somber  😭

laughing    😂
lol 😂
laughing crying 😂
tears   😂
joy 😂
happy   😂
laughing tears  😂
cry 😂
face with tears of joy  😂