word2vecfz

word2vecfz copied to clipboard

This is a modification of Yoav Goldberg's modification of word2vec! Yay for forking. Namely, this is a small change to the Dependency-based Word Embeddings software presented by (Levy and Goldberg, 2014).

I guess I'll call it word2vecfz

In this modification: - Removed the ability to perform multiple iterations over data - Removed the variation of word2vec - Added the ability to read from a BZ2 compressed corpus file.

This last change is the big one. Now the -train file must be bz2 compressed. It must be a single stream (so no cat'ing multiple bz2 files).

I did this because the corpus file can easily reach hundreds of gigabytes when using the 4B word corpora I've employed, but the compressed file will be maybe 20gb. Since I have strict quotas at University, this is my work around.

This is a modification of the word2vec software by Mikolov et.al, allowing:

• performing multiple iterations over the data.
• the use of arbitraty context features.
• dumping the context vectors at the end of the process.

This software was used in the paper "Dependency-Based Word Embeddings", Omer Levy and Yoav Goldberg, 2014.

The "main" binary is word2vecf. See README.word2vecf.txt for usage instructions.

Unlike the original word2vec program which is self-contained, the word2vecf program assumes some precomputations.

In particular, word2vecf DOES NOT handle vocabulary construction, and does not read an unprocessed input.

The expected files are: word_vocabulary: file mapping words (strings) to their counts context_vocabulary: file mapping contexts (strings) to their counts used for constructing the sampling table for the negative training. training_data: textual file of word-context pairs. each pair takes a seperate line. the format of a pair is " ", i.e. space delimited, where and are strings. if we want to prefer some contexts over the others, we should construct the training data to contain the bias.

(content below is the README.txt file of the original word2vec software)

Tools for computing distributed representtion of words

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 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
• 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.

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.

More information about the scripts is provided at https://code.google.com/p/word2vec/