More Statistical Machine Translation       


Artificial Intelligence Accomplishment | 2009

IBM researchers: Salim Roukos

Where the work was done: IBM T.J. Watson Research Center

What we accomplished: The field of statistical machine translation was revolutionized by casting the problem as a statistical communication model rather than the then-traditional approach of rule-based systems. The key challenges involved the development of a series of steps to estimate more complex translation models from earlier easier -- and cruder -- translation models. A sequence of five models was used to estimate a word alignment between the words of a source and a target sentence. These models are referred to in the scientific literature as "IBM Model through IBM Model 5." (Salim Roukos, pictured)

Related links: Wikipedia. Key Publications: (1) System Combination for Machine Translation of Spoken and Written Language; (2) A maximum entropy word aligner for Arabic-English machine translation

Image credit: IBM

BACK TO ARTIFICIAL INTELLIGENCE

BACK TO IBM RESEARCH ACCOMPLISHMENTS