SRILM - The SRI Language Modeling Toolkit

SRILM is a toolkit for building and applying statistical language models (LMs), primarily for use in speech recognition, statistical tagging and segmentation. It has been under development in the SRI Speech Technology and Research Laboratory since 1995. The toolkit has also greatly benefitted from its use and enhancements during the Johns Hopkins University/CLSP summer workshops in 1995, 1996, 1997, and 2002 (see history).

These pages and the software itself assume that you know what statistical language modeling is. To learn about language modeling we recommend the textbooks

• Speech and Language Processing by Dan Jurafsky and Jim Martin
• Foundations of Statistical Natural Language Processing by Chris Manning and Hinrich Schütze.

Chapter 6 in either book gives an excellent introduction to N-gram language modeling, which is the main type of LM supported by SRILM.

SRILM consists of the following components:

• A set of C++ class libraries implementing language models, supporting data stuctures and miscellaneous utility functions.
• A set of executable programs built on top of these libraries to perform standard tasks such as training LMs and testing them on data, tagging or segmenting text, etc.
• A collection of miscellaneous scripts facilitating minor related tasks.

SRILM runs on UNIX and Windows platforms.

Documentation

SRILM is still under development. The documentation in particular is work in progress. Best documented are the executable programs, scripts, and file formats, in the form of UNIX-style manual pages. The libraries are documented mostly in the source code. An overview of what the software can do and its design philosophy can be found in the paper "SRILM - An Extensible Language Modeling Toolkit", in Proc. Intl. Conf. Spoken Language Processing, Denver, Colorado, September 2002 (postscript, PDF). Links to other papers and tutorials, as well as frequently asked questions, are also given here.

Terms of Use

We are releasing SRILM mainly because we have received requests for the software from several researchers around the world. Therefore, there is no guarantee of support. SRILM can be downloaded free of charge under an "open source community license", meaning that you can use it freely for non-profit purposes, as long as you share any changes you make with the rest of the user community. For other uses please contact the author.

Mailing List

Exchange of information among SRILM users, as well as some level of technical support, is provided through the mailing list srilm-user@speech.sri.com. Check the mailing list archive for past contributions.

To subscribe and obtain more information, email the line "subscribe srilm-user" to majordomo@speech.sri.com.

SRILM is listed with the Natural Language Software Registry and Language Technology World.