eac / word2vec

Go library for performing computations in word2vec binary models

Home Page:https://www.sajari.com

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

word2vec

Build Status GoDoc

word2vec is a Go package which provides functions for querying word2vec models (see https://code.google.com/p/word2vec). Any binary word2vec model file can be loaded and queried.

Requirements

Installation

If you haven't setup Go before, you need to first set a GOPATH (see https://golang.org/doc/code.html#GOPATH).

To fetch and build the code:

$ go get github.com/sajari/word2vec/...

This will build the command line tools (in particular word-calc, word-server, word-client) into $GOPATH/bin (assumed to be in your PATH already).

Usage

word-calc

The word-calc tool is a quick way to perform basic word calculations on a word2vec model. For instance: vec(king) - vec(man) + vec(woman) would be equivalent to:

$ word-calc -model /path/to/model.bin -add king,woman -sub man

See word-calc -h for full more details. Note that word-calc first loads the model every time, and so can appear to be quite slow. Use word-server and word-client to get better performance when running multiple queries on the same model.

word-server and word-client

The word-server tool (see cmd/word-server) creates an HTTP server which wraps a word2vec model which can be queried from Go using a Client, or using the word-client tool (see cmd/word-client).

$ word-server -model /path/to/model.bin -listen localhost:1234

A simple code example using Client:

c := word2vec.Client{Addr: "localhost:1234"}

// Create an expression.
expr := word2vec.Expr{}
expr.Add(1, "king")
expr.Add(-1, "man")
expr.Add(1, "woman")

// Find the most similar result by cosine similarity.
matches, err := c.CosN(expr, 1)
if err != nil {
	log.Fatalf("error evaluating cosine similarity: %v", err)
}

API Example

Alternatively you can interact with a word2vec model directly in your code:

// Load the model from an io.Reader (i.e. a file).
model, err := word2vec.FromReader(r)
if err != nil {
	log.Fatalf("error loading model: %v", err)
}

// Create an expression.
expr := word2vec.Expr{}
expr.Add(1, "king")
expr.Add(-1, "man")
expr.Add(1, "woman")

// Find the most similar result by cosine similarity.
matches, err := model.CosN(expr, 1)
if err != nil {
	log.Fatalf("error evaluating cosine similarity: %v", err)
}

About

Go library for performing computations in word2vec binary models

https://www.sajari.com

License:MIT License


Languages

Language:Go 100.0%