geez
Semantle #22 in 285 guesses. My first guess had a similarity of 4.51. My first word in the top 1000 was at guess #19. My penultimate guess had a similarity of 19.19. semantle.novalis.org
Conversation
Age-restricted adult content. This content might not be appropriate for people under 18 years old. To view this media, you’ll need to log in to Twitter. Learn more
word2vec does something funny to adverbs, something like clustering them closer to the words they're modifying than feels entirely intuitive to me when i'm thinking about substitutability? i have to keep switching back and forth
1
7
friendly reminder that word2vec is not a measure of semantic similarity, really; closer to "do these words appear in similar places in similar sentences in similar newspaper articles"
Quote Tweet
Replying to @Christintweets and @QiaochuYuan
I learned yesterday from my friends that word2vec was trained on a corpus of google news articles 
so it was really “most likely to appear in the same news article” rather than synonyms as I initially misunderstood
so it was really “most likely to appear in the same news article” rather than synonyms as I initially misunderstood2
8
also, the image a few tweets up has a sensitive content warning on it so that you don't see it unless you want to b/c it's a spoiler; if you disabled these then i guess it won't work so uh, if that's you try re-enabling them ¯\_(ツ)_/¯
1
2
Replying to
Gave up after 95 (I know, lame)
I misunderstood the game and thought "nearby" words would be nearby synonyms, not that the words are associated with each other in common use
Maybe I'll try again tomorrow, but man, that is a hard game
1
3
I already had a word in the top 5 at this point but couldn't think of any more synonym-ish words
1
Replying to
I solved Semantle #22 in 311 guesses. My first guess had a similarity of 11.96. My first word in the top 1000 was at guess #41. My penultimate guess had a similarity of 19.85. semantle.novalis.org 💀
1
2
I learned yesterday from my friends that word2vec was trained on a corpus of google news articles 🤦🏻♀️ so it was really “most likely to appear in the same news article” rather than synonyms as I initially misunderstood
1
1
3
Show replies
Replying to
I did 100 guesses and got nowhere. I got kind of stuck in sentence joiners and ran out of ideas. Random craphshots landed completely cold. Are these join words just a trap that are always kinda close to anything? Would be cool if there was a better way to visualise the space.
1
1
i would avoid them personally, i think their similarity ratings are too noisy to do anything with but not sure
1