MALLET

MALLET: A Machine Learning for Language Toolkit

This page is out of date. Please visit http://mallet.cs.umass.edu.

MALLET is an integrated collection of Java code useful for statistical natural language processing, document classification, clustering, information extraction, and other machine learning applications to text.

It was written by Andrew McCallum, with contributions from several graduate students and staff, at University of Massachusetts Amherst, as well as contributions from Fernando Pereira, Ryan McDonald, and others.

Its development was funded by DARPA and the Airforce Research Laboratory (AFRL) under contract numbers F30602-00-2-0597 and F30602-01-2-0566, the National Science Foundation under grant EIA-9983215.

You might also be interested in other similar software packages for machine learning applied to text.

About the toolkit

The toolkit provides facilities for:

Recursively descending directories, finding text files.
Quite arbitrary pipelines of text processing steps.
Tokenizing a text file, according to arbitrary regular expressions.
Including N-grams among the tokens.
Creating real-valued feature vectors, and feature vector sequences.
Mapping strings to integers and back again, very efficiently.
Pruning vocabulary by information gain.
Building and manipulating feature vectors.
A general framework for optimization (based on "Numerical Recipes in C").
Several classification methods including naive Bayes, maximum entropy, Boosting, Winnow.
Maximum entropy classifier training is highly efficient, making use of Nocedal's "Limited-Memory BFGS", an efficient quasi-Newton optimization technique. It also handles arbitrary real-valued features.
A general framework for finite state transducers.
An implementation of Conditional Random Fields, also trained by Limited-Memory BFGS.
Writing models to disk.
Performing test-train splits.

Although many portions of the toolkit are very stable and usable, the toolkit as a whole is still in quite early stages of development. There is unfortunately almost no documentation at this point; it will come eventually. In the mean time, there are good examples of code for front-end functionality in directories named "examples", and in source files named "TUI.java".

The toolkit is Open Source Software, and is released under the Common Public License.

Citation

You are welcome to use the code under the terms of the licence for research or commercial purposes, however please acknowledge its use with a citation:

   McCallum, Andrew Kachites.  "MALLET: A Machine Learning for
   Language Toolkit."	http://www.cs.umass.edu/~mccallum/mallet.
   2002.

Here is a BiBTeX entry:

   @unpublished{McCallumMALLET,
      author = "Andrew Kachites McCallum",
      title = "MALLET: A Machine Learning for Language Toolkit",
      note = "http://www.cs.umass.edu/~mccallum/mallet",
      year = 2002}

Obtaining the Source

Source code for the library can be downloaded from this directory (sorry, first release not yet available, but will be soon). Different versions are indicated by eight digit sequences that indicate year, month and day. Thus, the most recent version is the one with the largest version number.

MALLET requires Java 1.4 or higher.

Unfortunately I do not have time to help all users with their compilation and usage problems. Feel free to send me mail asking for help, but please do not necessarily expect me to have time to help. Most appreciated are bug reports accompanied by fixes.