International Workshop on Spoken Language Translation (IWSLT) 2011

San Francisco, CA, USA
December 8-9, 2011

The MIT-LL/AFRL IWSLT-2011 MT System

A. Ryan Aminzadeh (1), Tim Anderson (2), Ray Slyh (2), Brian Ore (2), Eric Hansen (2), Wade Shen (3), Jennifer Drexler (3), Terry Gleason (3)

(1) Department of Defense, Washington, DC, USA
(2) Air Force Research Laboratory, Human Effectiveness Directorate, Wright-Patterson AFB, OH, USA
(3)MIT/Lincoln Laboratory, Human Language Technology Group, Lexington, MA, USA

This paper describes the MIT-LL/AFRL statistical MT system and the improvements that were developed during the IWSLT 2011 evaluation campaign. As part of these efforts, we experimented with a number of extensions to the standard phrase-based model that improve performance on the Arabic to English and English to French TED-talk translation tasks. We also applied our existing ASR system to the TED-talk lecture ASR task.
   We discuss the architecture of the MIT-LL/AFRL MT system, improvements over our 2010 system, and experiments we ran during the IWSLT-2011 evaluation. Specifically, we focus on 1) speech recognition for lecture-like data, 2) cross-domain translation using MAP adaptation, and 3) improved Arabic morphology for MT preprocessing.

Full Paper

Bibliographic reference.  Aminzadeh, A. Ryan / Anderson, Tim / Slyh, Ray / Ore, Brian / Hansen, Eric / Shen, Wade / Drexler, Jennifer / Gleason, Terry (2011): "The MIT-LL/AFRL IWSLT-2011 MT system", In IWSLT-2011, 34-40.