The Stanford NLP (Natural Language Processing) Group

Statistical Machine Translation Project
Homepage

Overview

The project aims to develop statistical translation models for natural language. The utility of the developed TM will be evaluated on different language pairs, including English-French, English-Bulgarian, and (possibly) English-Japanese. The main goals are

Investigating on new translation models, within and outside the paradigm for noisy channel translation proposed by Brown et al (1993).
Making use of syntactic knowledge of language and other linguistic resources.
Investigating on the decoding problem for noisy channel translation.

People

Professors:
- Chris Manning
Ph.D. Students:
- Tolga Ilhan
- Kristina Toutanova

Related Work

The Mathematics of Statistical Machine Translation P. Brown, S. Della Pietra, V. Della Pietra, R. Mercer, in Computational Linguistics 19 (2), 1993.
A Syntax-Based Statistical Translation Model K. Yamada and K. Knight , in Proc. of the Conference of the Association for Computational Linguistics (ACL), 2001

Fast Decoding and Optimal Decoding for Machine Translation U. Germann, M. Jahr, K. Knight, D. Marcu, and K. Yamada, in Proc. of the Conference of the Association for Computational Linguistics (ACL), 2001

A Comparison of Alignment Models for Statistical Machine Translation Franz Josef Och, Hermann Ney. in COLING '00: The 18th Int. Conf. on Computational Linguistics, pp. 1086-1090, Saarbr�cken, Germany, July 2000.

The Workshop on Statistical Machine Translation at JHU 99

Progress

Implemented Model 1 and HMM in Perl
Experimented with incorporating POS tag information into the translation models
Implemented Model 1 in Java
Experimented with syntax-driven word reordering

Overview

People

Related Work

Progress