vane
/
word2vec
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

removed some specific information from README, as I just changed the scritps

This commit is contained in:
tmikolov 2013-08-01 20:12:56 +00:00
parent 402ae80561
commit ada9ca3db6
1 changed files with 13 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
README.txt
View File

@ -1,29 +1,21 @@
Tools for computing distributed representtion of words
------------------------------------------------------
We provide an implementation of the Continuous Bag-of-Words (CBOW) and the Skip-gram model (SG).
We provide an implementation of the Continuous Bag-of-Words (CBOW) and the Skip-gram model (SG), as well as several demo scripts.
Given a text corpus, the word2vec program learns a vector for every word using the Continuous
Bag-of-Words or the Skip-Gram model. The user needs to specify the following:
Given a text corpus, the word2vec tool learns a vector for every word in the vocabulary using the Continuous
Bag-of-Words or the Skip-Gram neural network architectures. The user should to specify the following:
- desired vector dimensionality
- the size of the context window for either the Skip-Gram or the Continuous Bag-of-Words model
- Whether hierarchical sampling is used
- Whether negative sampling is used, and if so, how many negative samples should be used
- A threshold for downsampling frequent words
- Number of threads to use
- Whether to save the vectors in a text format or a binary format
- training algorithm: hierarchical softmax and / or negative sampling
- threshold for downsampling the frequent words
- number of threads to use
- the format of the output word vector file (text or binary)
Usually, the other hyper-parameters such as the learning rate do not need to be tuned for different training sets.
Thus the programs require a very modest number of parameter. In particular, learning rates
need not be selected.
The script demo-word.sh downloads a small (100MB) text corpus from the web, and trains a small word vector model. After the training
is finished, the user can interactively explore the similarity of the words.
The file demo-word.sh downloads a small (100MB) text corpus, and trains a 200-dimensional CBOW model
with a window of size 5, negative sampling with 5 negative samples, a downsampling of 1e-3, 12 threads, and binary files.
./word2vec -train text8 -output vectors.bin -cbow 1 -size 200 -window 5 -negative 5 -hs 0 -sample 1e-3 -threads 12 -binary 1
Then, to evaluate the fidelity of our vectors, we can run the command, which will run
a battery of tests on the vectors to determine their fidelity. The tests evaluate
the vectors' ability to perform linear analogies.
./distance vectors.bin
More information about the scripts is provided at https://code.google.com/p/word2vec/