International Workshop on Spoken Language Translation (IWSLT) 2012

Hong Kong
December 6-7, 2012

The MIT-LL/AFRL IWSLT-2012 MT System

Jennifer Drexler (1), Wade Shen (1), Terry (1), Tim Anderson (2), Raymond Slyh (2), Brian Ore (2), Eric Hansen (2)

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

This paper describes the MIT-LL/AFRL statistical MT system and the improvements that were developed during the IWSLT 2012 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 task. We also applied our existing ASR system to the TED-talk lecture ASR task, and combined our ASR and MT systems for the TED-talk SLT task.
    We discuss the architecture of the MIT-LL/AFRL MT system, improvements over our 2011 system, and experiments we ran during the IWSLT-2012 evaluation. Specifically, we focus on 1) cross-domain translation using MAP adaptation, 2) cross-entropy filtering of MT training data, and 3) improved Arabic morphology for MT preprocessing

Full Paper   

Bibliographic reference.  Drexler, Jennifer / Shen, Wade / , Terry / Anderson, Tim / Slyh, Raymond / Ore, Brian / Hansen, Eric (2012): "The MIT-LL/AFRL IWSLT-2012 MT system", In IWSLT-2012, 109-116.