Skip to content

carschno/JFastText

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Build Status

Table of Contents

Introduction

JFastText is a Java wrapper for Facebook's fastText, a library for efficient learning of word embeddings and fast sentence classification. The JNI interface is built using javacpp.

The library provides full fastText's command line interface. It also provides the API for loading trained model from file to do label prediction in memory. Model training and quantization are supported via the command line interface.

JFastText is ideal for building fast text classifiers in Java.

Maven Dependency

<dependency>
  <groupId>io.github.carschno</groupId>
  <artifactId>jfasttext</artifactId>
  <version>0.9.1</version>
</dependency>

The Jar package on Maven Central is bundled with precompiled fastText library for Windows, Linux and MacOSX 64bit.

Windows and Mac OS X

Currently, the Maven dependency only contains binaries for Linux (64 bit), not for Windows or Mac OS X. In order to use JFastText for Windows or Mac OS X (or any other system), you need to build it yourself (see below).

Building

C++ compiler (g++ on Mac/Linux or cl.exe on Windows) is required to compile fastText's code.

git clone --recursive https://github.com/carschno/JFastText
cd JFastText
git submodule init
git submodule update
mvn package

Building on Windows

The (automatic) build seems to fail on some Windows systems/C++ compilers. See this issue:

I used MS's developer tools, not the full-blown Visual Studio. If I run cl directly, the compilation fails with the same error.

I was able to build on Windows by changing the call to cl.exe and running it outside the Maven build. I changed one parameter in the call to cl: I use /MT (whereas Maven uses /MD). Bundling the generated DLLs works fine.

Quick Application - Language Identification

JFastText can use FastText's pretrained models directly. Language identification models can be downloaded here. In this quick example, we will use the quantized model which is super small and a bit less accurate than the original model.

$ wget -q https://s3-us-west-1.amazonaws.com/fasttext-vectors/supervised_models/lid.176.ftz \
    && { echo "This is English"; echo "Xin chào"; echo "Привет"; } \
    | java -jar target/jfasttext-*-jar-with-dependencies.jar predict lid.176.ftz -
__label__en
__label__vi
__label__ru

Detailed Examples

Examples on how to use JFastText can be found at examples/api and examples/cmd.

API

Initialization

import com.github.jfasttext.JFastText;
...
JFastText jft = new JFastText();

Word embedding learning

jft.runCmd(new String[] {
        "skipgram",
        "-input", "src/test/resources/data/unlabeled_data.txt",
        "-output", "src/test/resources/models/skipgram.model",
        "-bucket", "100",
        "-minCount", "1"
});

Text classification

// Train supervised model
jft.runCmd(new String[] {
        "supervised",
        "-input", "src/test/resources/data/labeled_data.txt",
        "-output", "src/test/resources/models/supervised.model"
});

// Load model from file
jft.loadModel("src/test/resources/models/supervised.model.bin");

// Do label prediction
String text = "What is the most popular sport in the US ?";
JFastText.ProbLabel probLabel = jft.predictProba(text);
System.out.printf("\nThe label of '%s' is '%s' with probability %f\n",
        text, probLabel.label, Math.exp(probLabel.logProb));

FastText's Command Line

FastText's command line interface can be accessed as follows:

$ java -jar target/jfasttext-*-jar-with-dependencies.jar
usage: fasttext <command> <args>

The commands supported by fasttext are:

  supervised              train a supervised classifier
  quantize                quantize a model to reduce the memory usage
  test                    evaluate a supervised classifier
  predict                 predict most likely labels
  predict-prob            predict most likely labels with probabilities
  skipgram                train a skipgram model
  cbow                    train a cbow model
  print-word-vectors      print word vectors given a trained model
  print-sentence-vectors  print sentence vectors given a trained model
  print-ngrams            print ngrams given a trained model and word
  nn                      query for nearest neighbors
  analogies               query for analogies
  dump                    dump arguments,dictionary,input/output vectors

For example:

$ java -jar target/jfasttext-*-jar-with-dependencies.jar quantize -h

License

BSD

References

(From fastText's references)

Please cite 1 if using this code for learning word representations or 2 if using for text classification.

Enriching Word Vectors with Subword Information

[1] P. Bojanowski*, E. Grave*, A. Joulin, T. Mikolov, Enriching Word Vectors with Subword Information

@article{bojanowski2016enriching,
  title={Enriching Word Vectors with Subword Information},
  author={Bojanowski, Piotr and Grave, Edouard and Joulin, Armand and Mikolov, Tomas},
  journal={arXiv preprint arXiv:1607.04606},
  year={2016}
}

Bag of Tricks for Efficient Text Classification

[2] A. Joulin, E. Grave, P. Bojanowski, T. Mikolov, Bag of Tricks for Efficient Text Classification

@article{joulin2016bag,
  title={Bag of Tricks for Efficient Text Classification},
  author={Joulin, Armand and Grave, Edouard and Bojanowski, Piotr and Mikolov, Tomas},
  journal={arXiv preprint arXiv:1607.01759},
  year={2016}
}

FastText.zip: Compressing text classification models

[3] A. Joulin, E. Grave, P. Bojanowski, M. Douze, H. Jégou, T. Mikolov, FastText.zip: Compressing text classification models

@article{joulin2016fasttext,
  title={FastText.zip: Compressing text classification models},
  author={Joulin, Armand and Grave, Edouard and Bojanowski, Piotr and Douze, Matthijs and J{\'e}gou, H{\'e}rve and Mikolov, Tomas},
  journal={arXiv preprint arXiv:1612.03651},
  year={2016}
}

(* These authors contributed equally.)

About

Java interface for fastText

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Java 77.0%
  • C++ 23.0%