Word2Bits
Languages and libraries
word embeddings library
SHOW HN2018
How HN took it
13 positive, 25 neutral or asking, 0 mixed, 3 critical, out of 41 comments about the project. Thread score +0.25.
Idea: 5 for, 2 againstPerformance: 3 for, 0 againstPrivacy: 0 for, 1 againstDocs: 1 for, 0 againstExecution: 1 for, 0 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Useful for memory-constrained and client-side applications
From 2 comments
This is great and echoes a recent fascination for me. One application of compact word embeddings is that if they're small enough you can ship a whole model to the user in a web…
dougb5 on HNcan this be done with fasttext as well? Word2bits is definitely great for memory-constrained applications but for server use memory isn't as much a constraint (there's a direct…
Jack000 on HN
Interesting results beating Word2Vec on SQuAD
From 1 comment
Very cool that this beats Word2Vec on SQuAD! I wonder if the current state of the art models are using the standard GloVe word vecs and might see improvement from this. Tbh I…
rococode on HN
Objections
Vectors may carry gender and other biases from training data
From 1 comment
It's interesting and slightly uncomfortable that the illustration for similar words uses the word "man" as an example, given the gender biases that result from learning word…
rspeer on HN
Dimensions appear to be noise and hard to interpret
From 1 comment
Definitely tried to figure out if the dimensions mean anything -- as far as I can tell they don't really mean much :(
maxlam on HN
Requests
Test debiasing methods on quantized vectors
From 2 comments
It's interesting and slightly uncomfortable that the illustration for similar words uses the word "man" as an example, given the gender biases that result from learning word…
rspeer on HNThere may be a way to build it into the loss function so that it happens before the quantization, right? (and holy crap, look how fast the HN conservatives are getting to my…
rspeer on HN
Add fastText-style out-of-vocabulary handling
From 1 comment
can this be done with fasttext as well? Word2bits is definitely great for memory-constrained applications but for server use memory isn't as much a constraint (there's a direct…
Jack000 on HN
Questions
How to read the graphs
From 1 comment
Can someone explain how to read the graph?
wodenokoto on HN
How many quantization values were tried
From 1 comment
Very cool that this beats Word2Vec on SQuAD! I wonder if the current state of the art models are using the standard GloVe word vecs and might see improvement from this. Tbh I…
rococode on HN
Alternatives named
word2vec and GloVe
From 2 comments
Very cool that this beats Word2Vec on SQuAD! I wonder if the current state of the art models are using the standard GloVe word vecs and might see improvement from this. Tbh I…
rococode on HNIt's interesting and slightly uncomfortable that the illustration for similar words uses the word "man" as an example, given the gender biases that result from learning word…
rspeer on HN
fastText
From 1 comment
can this be done with fasttext as well? Word2bits is definitely great for memory-constrained applications but for server use memory isn't as much a constraint (there's a direct…
Jack000 on HN