Git Product home page Git Product logo

word2vec-pytorch's Introduction

Word2vec in Pytorch

This repo has learnt a lot from this repo

This repo implements the SkipGram model with negative sampling of Word2vec by Mikolov.

Tricks below are also implemented:

  • subsampling
  • negative sampling with pow weight decay
  • learning rate decay

Requirements

  • PyTorch >= 0.4.1
  • Gensim >= 3.6.0 (for testing only)

Fast run

To quickly run the train model, just run

python train.py

which uses a Chinese corpus to train the Word2vec model. There is another toy corpus in English you can use located in data/trainset.txt, which is actually a "Jane Eyre" novel.

Issues and PRs are welcomed!

Reference

[1] Mikolov T, Sutskever I, Chen K, et al. Distributed representations of words and phrases and their compositionality[C]//Advances in neural information processing systems. 2013: 3111-3119.

word2vec-pytorch's People

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar

word2vec-pytorch's Issues

subsampling

in def gen_vocab(self) we select the vocab that have number of freq >=self.min_count like this:

  1. `vocab, word2id, id2word = {}, {}, {}
  2.     index = 0
    
  3.     for item_id, freq in vocab_freq_dict.items():
    
  4.         if freq < self.min_count:
    
  5.             continue
    
  6.         vocab[item_id] = freq
    
  7.         word2id[item_id] = index
    
  8.         id2word[index] = item_id
    
  9.         index += 1
    
  10.     return vocab, word2id, id2word, total_word_count, total_sent_count`
    

can you please clarify this function (def gen_subsample_table(self))?

`

  1. def gen_subsample_table(self):
  2.     """
    
  3.     sub sampling rate, higher than that would be sub sampled using
    
  4.         the word2vec paper using:    p(w_i) = 1 - sqrt(sub_sampling / freq)
    
  5.         the word2vec code using:     p(w_i) = 1 - (sqrt(sub_sampling / freq) + sub_sampling / freq)
    
  6.     we use word2vec code sub sampling method here.
    
  7.     :return: {word_id: sample_score}
    
  8.     """
    
  9.     def sub_sampling(_freq):
    
  10.         return (self.sub_sampling_t / 1.0 / _freq) ** 0.5 + self.sub_sampling_t / 1.0 / _freq
    
  11.     # word freq count to word freq ratio
    
  12.     sub_sample_tbl = {item: freq / 1.0 / self.total_word_count
    
  13.                       for item, freq in self.vocab.items()
    
  14.                       if freq / 1.0 / self.total_word_count > self.sub_sampling_t}
    
  15.     # freq to score
    
  16.     sub_sample_tbl = {item: sub_sampling(_freq) for item, _freq in sub_sample_tbl.items()}
    
  17.     # word to id
    
  18.     sub_sample_tbl = {self.word2id[i]: j for i, j in sub_sample_tbl.items() if j < 1}
    
  19.     return sub_sample_tbl
    

`
line 9
9. def sub_sampling(_freq): it looks like it returns ( p(w_i) = (sqrt(sub_sampling / freq) + sub_sampling / freq) ) not ( p(w_i) = 1 - (sqrt(sub_sampling / freq) + sub_sampling / freq) ) right?

why this line ?
14. if freq / 1.0 / self.total_word_count > self.sub_sampling_t}
if we before used
4. if freq < self.min_count: in the def gen_vocab(self) function in the first part of the question

what is the meaning of this line?

  1.     sub_sample_tbl = {self.word2id[i]: j for i, j in sub_sample_tbl.items() if j < 1}
    

thank you!

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.