GA-FLM is a genetic algorithms program for automatically learning factored language model structures. It is used as an extension to the factored language model programs in the SRI Language Modeling toolkit.

The program takes as input some training/development text files and some parameter files that specify the type of genetic algorithms and factored language model desired by the user. It then uses standard genetic algorithms search to build a population of factored language models and optimizes for their development set perplexity.

For more information, refer to the coling2004.pdf paper in the doc directory.

GA-FLM -- Genetic Algorithms for Learning Factored Language Model Structure Copyright (c) 2004 University of Washington

Written by Kevin Duh and Sonia Parandekar

NO WARRANTY THE PROGRAM IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED INCLUDING, WITHOUT LIMITATION, ANY WARRANTIES OR CONDITIONS OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Each Recipient is solely responsible for determining the appropriateness of using the Program and assumes all risks associated with such use, including but not limited to the risks and costs of program errors, compliance with applicable laws, damage to or loss of data, programs or equipment, and unavailability or interruption of operations.

DISCLAIMER OF LIABILITY THE UNIVERSITY OF WASHINGTON, KEVIN DUH AND SONIA PARANDEKAR SHALL NOT HAVE ANY LIABILITY FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING WITHOUT LIMITATION LOST PROFITS), HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THE PROGRAM, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES."

For more information, bug reports, fixes, contact: Kevin Duh Dept. of Electrical Engineering, Paul Allen Center University of Washington Seattle, WA 98195 USA duh@ee.washington.edu (http://ssli.ee.washington.edu/people/duh/ Homepage) --

• Make sure you have the SRI Language Modeling Toolkit version 1.4.1 or above. The two executables that are needed in that package are "fngram-count" and "fngram", which builds and tests factored language models. The toolkit can be downloaded at: http://www.speech.sri.com/projects/srilm

• To install, simply type "make" in this directory to create the executable "ga-flm"

There are two ways to run the program, either by running the executable "ga-flm" directly or running the "repeat-ga-flm.pl" Perl wrapper function.

ga-flm [-g ga_param_file] [-f factoredLM_param_file] [-s seedfile]

perl repeat-ga-flm.pl [num] [-g ga_param_file] [-f factoredLM_param_file] [-s seedfile]

Refer to the manual.pdf in the doc directory for more detailed explanation.

TODO

• GA-FLM is the work of several people. The algorithm for automatically learning factored language model structure was developed with Katrin Kirchhoff. Jeff Bilmes gave invaluable support on the SRILM factored language model programs. Sonia Parandekar wrote almost all of the basic genetic algorithms code.

• Published research using GA-FLM may cite the following paper:

Kevin Duh and Katrin Kirchhoff, "Automatic Learning of Language Model Structure", Proc. of the 20th International Conference on Computational Linguistics (COLING-2004), Geneva, Switzerland, August 2004.