Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: Language games – simple games made with word vectors

github.com

Post reply on HN
Print view View original

11–20 of 22 posts

Re: Show HN: Language games – simple games made with word vectors

#11

Post by dcsan »

Der_Einzige wrote: post #3
r34 wrote: post #2

I wonder if there are any good sets of word vectors in languages other than English (I'm specifically interested in Polish). Same with any other language models and NLP related stuff. It's hard to find anything valuable.

Yeah! Fasttext has some models in polish which I can convert into the magnitude format for you! File an issue for it and I'll get it done

Yes I would look to fasttext first, they have a huge library of vectors here: https://fasttext.cc/docs/en/crawl-vectors.html

Actually facebook research are pretty good at gathering things, their ParlAI project also gathers together other sources for conversational stuff.

Re: Show HN: Language games – simple games made with word vectors

#13

Post by dcsan »

btw what did you like most about using the magnitude format? lazy loading seems good, depending on if it takes time to load the sqlite DB at each startup. That's one of the major pains of working on anything with word embeddings is that server reloads / hot restarts take forever.

The fastText vectors (see below) also use subword embeddings so I think they had potentially better results. I used FT for some Chinese stuff and I think it worked better for that since chinese characters are so much more important than latin scripts.

Seems magnitude python API also has some other features like POS tagging, I wonder how that compares to say spaCy.

Re: Show HN: Language games – simple games made with word vectors

#14

Post by johnnyheineken »

Nice project!

I did something similar when I was still at school. Scripts which emulated Codenames [1], and providing hints for given set of words [2].

I believe that is possible extension to your project - emulate Codenames games:)

[1] https://boardgamegeek.com/boardgame/178900/codenames

[2] https://github.com/johnnyheineken/codenames-AI

Re: Show HN: Language games – simple games made with word vectors

#15

Post by SamBam »

I'm failing to understand game 4:

    Player 0 it is your turn!
    The word you are trying to match in meaning is:
    antepenultimate
    Your char list is:
    ['i', 'g', 'j', 'u', 'b']
    Please input a valid word made from some or all of the provided characters
What's the right answer?

Re: Show HN: Language games – simple games made with word vectors

#16

Post by chosenPixel »

SamBam wrote: post #15

I'm failing to understand game 4: Player 0 it is your turn! The word you are trying to match in meaning is: antepenultimate Your char list is: ['i', 'g', 'j', 'u', 'b'] Please input a valid word made from some or all of the provided characters What's the right answer?

'bug' would be one i'd assume

Re: Show HN: Language games – simple games made with word vectors

#17

Post by SamBam »

chosenPixel wrote: post #16
SamBam wrote: post #15

I'm failing to understand game 4: Player 0 it is your turn! The word you are trying to match in meaning is: antepenultimate Your char list is: ['i', 'g', 'j', 'u', 'b'] Please input a valid word made from some or all of the provided characters What's the right answer?

'bug' would be one i'd assume

But how does "bug" match the meaning of "antepenultimate" in any way? I must be missing the point of the game.

Is it just "words that vaguely, extremely distantly, have some relation, as determined by the neural net"? And your job is to work out which extremely-distant word is slightly closer? (I noticed "gubi" scored very, very slightly more points than "bug" in the demo.

Re: Show HN: Language games – simple games made with word vectors

#18

Post by ggggtez »

SamBam wrote: post #15

I'm failing to understand game 4: Player 0 it is your turn! The word you are trying to match in meaning is: antepenultimate Your char list is: ['i', 'g', 'j', 'u', 'b'] Please input a valid word made from some or all of the provided characters What's the right answer?

Importantly: there was no "right answer" in mind when it created the question. The letters were random, as was the word.

It calculates the distance in vector space from your word to the given word.

I think it helps if you understand what Word2Vec is, as then it becomes clear what is going on.

Re: Show HN: Language games – simple games made with word vectors

#19

Post by ggggtez »

SamBam wrote: post #17
chosenPixel wrote: post #16

Earlier quoted context omitted.

'bug' would be one i'd assume

But how does "bug" match the meaning of "antepenultimate" in any way? I must be missing the point of the game. Is it just "words that vaguely, extremely distantly, have some relation, as determined by the neural net"? And your job is to work out which extremely-distant word is slightly closer? (I noticed "gubi" scored very, very slightly more points than "bug" in the demo.

You are basically correct. https://en.wikipedia.org/wiki/Word2vec

Essentially, all words are embedded in an N-dimentional space. Then you can simply measure the distance between words in the word-cloud. The exact method of how it decided what coordinates to assign to each word in the cloud isn't important to the game.

Re: Show HN: Language games – simple games made with word vectors

#20

Post by alew1 »

johnnyheineken wrote: post #14

Nice project! I did something similar when I was still at school. Scripts which emulated Codenames [1], and providing hints for given set of words [2]. I believe that is possible extension to your project - emulate Codenames games:) [1] https://boardgamegeek.com/boardgame/178900/codenames [2] https://github.com/johnnyheineken/codenames-AI

I’ve seen your Codenames work project — it’s neat!

I’ve also played around with word vector games — Robot Mind Meld [1] has you and a robot working together to converge on the same word.

[1] http://robotmindmeld.com

Post reply on HN