We describe a machine translation approach being designed at HKUST to integrate semantic processing into statistical machine translation, beginning with entity and word sense disambiguation. We show how integrating the semantic modules consistently improves translation quality across several data sets. We report results on five different IWSLT 2006 speech translation tasks, representing HKUST’s first participation in the IWSLT spoken language translation evaluation campaign. We translated both read and spontaneous speech transcriptions from Chinese to English, achieving reasonable performance despite the fact that our system is essentially text-based and therefore not designed and tuned to tackle the challenges of speech translation. We also find that the system achieves reasonable results on a wide range of languages, by evaluating on read speech transcriptions from Arabic, Italian, and Japanese into English. 1.

Abstract. In this article a fast research into applications of Word Sense Disambiguation (WSD) in the machine translation (MT) is reported. The topic was chosen on author’s own experiences with MT output. It suffers from many problems. Faulted syntax constraints, not translated words and inappropriate lexical equivalent choice seems to be the most disturbing. The first problem usually occurs if the constraint should be used on distant words. Since the equivalent choice in MT is a topic of my PhD. study, I focus on the last mentioned problem. I hope that integrating WSD module into the right place in phrasal SMT will help. I summarize reported results that I have taken up with. Some of them demonstrate that MT benefits from WSD module, others show that the influence may even be negative. The main reported idea is that WSD method may improve MT when integrated properly.