GISK project page

Grammatical Inference with String Kernels

This page describes a pump priming project supported by the PASCAL Network of Excellence. The purpose of the project is to explore a new family of grammatical inference algorithms, based on the use of string kernels. These algorithms are capable of efficiently learning some languages that are context sensitive, including many linguistically interesting examples of mildly context sensitive languages. The project started on November 1st 2005, and finished at the end of October 2006. It is a collaboration between Royal Holloway and EURISE.

People

Publications

  1. Planar Languages and Learnability, Alexander Clark, Christophe Costa Florencio, Chris Watkins and Mariette Serayet, International Conference on Grammatical Inference (ICGI), 2006, September, Tokyo. paper in pdf
  2. Languages as Hyperplanes: grammatical inference with string kernels, Alexander Clark, Christophe Costa Florencio and Chris Watkins, ECML 2006, September, Berlin. paper in pdf . The slides for Chris Watkins' presentation are here
A long presentation that describes the overall project is available here

Data and code

We have prepared random data sets for a variety of context sensitive, context free and regular languages. These are as described in the ECML paper above, and there is some documentation included.
  1. Tar file of the datasets
  2. Tar file of the matlab code for doing grammatical inference with string kernels.

Last update: 8 November 2006, Website maintained by alexc@cs.rhul.ac.uk