Demand for Machine Translation technology is taking off as global information exchange proliferates on the internet. This has spurred a world-wide resurgence of machine translation research centered around data-driven techniques: painstakingly hand-crafted linguistic approaches strain to cope with the wide variety of language pairs and specialized jargon presented by the internet. The advent of large parallel text collections, increased computing power, and reliable automatic evaluation metrics heralds an exciting new era for high-quality machine translation. In this lecture, we discuss the foundations of statistical machine translation and automatic evaluation of translation quality.

Kishore Papineni is a Research Staff Member at the IBM T. J. Watson Research Center. He graduated from Rice University in 1995 with a Ph. D. in Electrical Engineering in the area of systems and control theory. For his Ph.D. he studied kernel representation of linear systems, H-infinity optimal control theory, and singular linear quadratic Gaussian control theory. He joined IBM Research in 1995 to work on natural language processing. He has worked on natural language understanding and dialog management. He is currently leading a team on statistical machine translation. He is interested in mathematical models applicable to natural language processing, with emphasis on exponential models.