“Segmental Conditional Random Fields for Speech Recognition”
Dr. Patrick Nguyen, Google Research
Sponsored by the Dallas Chapter of the IEEE Signal Processing Society
Novel techniques in speech recognition are often hampered by the long road that must be followed to turn them into fully functional systems capable of competing with the state-of-the-art. In this work we explore the use of segmental conditional random fields as an integrating technology that can augment the best conventional systems with information from novel scientific approaches. Segmental conditional random fields provide a principled, flexible framework to express a new class of claims that are either impossible or cumbersome to integrate in current speech recognition systems based on hidden Markov models. We illustrate the approach with work done at Microsoft and applied at a Johns Hopkins University workshop in which we find that the SCRF framework is able to appropriately weight different information sources. (Joint work with Geoff Zweig at Microsoft Research.)
Patrick Nguyen is a research scientist in Google Research. Prior to joining Google he was with Microsoft Research from 2004 to 2010. Prior to that, he worked at the Panasonic Speech Technology Laboratory for four years. In 1998 he started a company that provided a market-maker platform for real-time foreign exchange trading. He received his doctorate from the Swiss Federal Institute for Technology in 2002. His area of expertise revolves around statistical processing human language, and in particular speech recognition. He participated in NIST competitions, notably contributing to the winning machine translation competition in English to Chinese in 2008. He has co-authored more than a dozen granted patents and numerous publications. He serves as a reviewer for all major speech and natural language research publications, and he is on the organizing committee of ASRU 2011.
[an error occurred while processing this directive]