Publications

Advanced search

Abstract

Jesús Tomás, Juan M. Vilar, Francisco Casacuberta. The ITI Statistical Machine Translation System. Proceedings of the TC-Star Speech to Speech Translation Workshop, 2006. pp. 49-55.

We describe one translation system currently under development at ITI. One of its major features is the use of a phrase based approach: he elementary units of translation are sequences of consecutive words thereby improving the performance in comparison to single word pproaches. The use of a monotonicity assumption makes search feasible. In this work, different methods for the selection of the egments, the training of the parameters and search algorithms are discussed. They are experimentally assessed.