Mei-Yuh Hwang, Chris Quirk
We will be sharing progmatic insights on building a successful Mandarin SR/MT system. Particularly comparison will be presented using the most successful techniques. We will also pay attention to topics specific to the Chinese spoken and/or written language. Corpora available in U.S. and in Asia will be discussed.
For translation, we will be presenting the core technologies first, both syntactic and statistic. Then we will briefly present translation of western languages, to be compared with Chinese-to-English. We will conclude our MT efforts using system combination.
Mei-Yuh got her Ph.D. in Computer Science at Carnegie Mellon University (CMU) in 1994. She has been working in the area of speech recognition for two decades. She was one of the main contributors of Sphinx-II speech recogizer (SR) at CMU. Her team won numerous top performance among DARPA sponsored SR evaluations. She joined Microsoft in 1994, University of Washington in 2004-2007 where she led and built the best Mandarin SR for the DARPA GALE project. She is currently back to Microsoft, transfering her expertize to machine translation.
After studying Computer Science and Mathematics at Carnegie Mellon University, Chris joined Microsoft in 2000, and started work with the Natural Language Processing group in 2001. Currently Chris is working on machine translation, exploring ways to combine traditional linguistic methods with recent statistical advances. In addition, Chris have spent some time investigating automated MT evaluation and translation confidence scoring. Most recently Chris have done some investigations into applying effective statistical machine translation techniques to monolingual problems.