Learning Computational Grammars

LCG is a EU-funded TMR project whose objective is to apply a variety of machine learning techniques to the common task of learning NL syntax and evaluating their potential with regard to NLP applications. Having been launched in April 1998, the project should terminate in September, 2001.

To further focus the network, initial research has been directed at one sub-area of syntax - noun phrase structure - on the basis of English language test materials. The machine methods currently applied include connectionist learning, instance-based learning, genetic algorithms, maximum enthropy and explanation-based learning. ISSCO's involvement in the LCG network builds on the institute's experience in corpus-based research and evaluation. It has provided linguistic analysis and evaluation of some of the machine learning results produced by the other network participants, together with complementing work on human language learning.

The members of the LCG network are:

Further information about the project and demos are available at its official web site: http://lcg-www.uia.ac.be/lcg/

Contact: Adelina Ivanova Adelina.Ivanova@issco.unige.ch

[ISSCO's logo]Back to ISSCO's home page

Sandra Manzi
Last modified: Thu Aug 31 15:33:49 MET DST 2000