UMIACS Computational Linguistics Colloquium, February 25, 1998

Advances in Empirical Methods for Translation Model Construction


I. Dan Melamed


University of Pennsylvania


UMIACS Computational Linguistics Colloquium

February 25, 1998, 4pm, AVW Room 4406


The quality of available translation models, statistical or not, is a major limiting factor in the performance of many applications, such as cross-language information retrieval. In this talk, I will present recent advances in automatic methods for translation model construction that can either supplement or replace manual methods. Major advances include the ability to handle arbitrarily noisy training data, the ability to incorporate various types of pre-existing knowledge, and the ability to account for non-compositional "phrasal" translation.


For the colloquium series schedule, see the UMD Computational Linguistics Colloquium Series web page at http://umiacs.umd.edu/~resnik/cl_colloquium/. If you are interested in meeting with the speaker, please contact Mari Broman Olsen (molsen@umiacs.umd.edu) or Philip Resnik (resnik@umiacs.umd.edu).