Mei-Yuh Hwang
Affiliate Professor at EE Department
University of Washington

Mobvoi is hiring!

Mei-Yuh is the director at Mobvoi AI Lab, in Redmond WA. Mobvoi makes speech-enabled smart IoT devices, from hardware to software, all in-house.

Mei-Yuh received her PhD in Computer Science from Carnegie Mellon University in December 1993, specializing in speech recongition. After graduation, she worked for Microsoft through 2004 in speech research and products. From 2004-2008 she joined University of Washington and led the GALE Mandarin speech recognition project. In 2008 she returned to Microsoft Research and was one of the core scientists behind Bing Translator. From 2012-2015, she led speech recognition and language understanding for non-English versions of Cortana in China, until she returned to U.S. in 2016, leading Mobvoi AI Lab. Over the past two decades, Mei-Yuh has made many significant contributions to spoken language processing, both in academia and in industry.

Speech recognition

Mei-Yuh was the first to propose Markov state clustering based on decision trees for continuous speech recognition. The idea of shared states (or senones as Mei-Yuh named it in 1992) has been widely adopted since its inception. Although CTC end-to-end speech recognition is gaining popularity, most commercial speech recognizers continue to use senones, including the popular open-source toolkit Kaldi. Her IEEE paper was a gold reference guide on building a strong Mandarin speech recognizer.

Skype speech-to-speech translator

Mei-Yuh implemented a map-reduce training infrastructure for Microsoft Translator, designed and implemented Bing Translation Hub that is still actively used by small business today. After the successful launch of non-English Cortana, Mei-Yuh further led her team to adapt Cortana speech recognizers to spontaneous conversation speech in 2014, for Skype speech-to-speech translation.

Spoken language understanding

To deliver non-English Cortana without human annotated data, Mei-Yuh designed an adapted translation algorithm which offered both paraphrasing and generalization capabilities with required slot tags. The protoype model was further improved via iterative data augmentation using RNN. The impressive success of Chinese Cortana gained much attention within China, and sparked the development of personal assistants and AI across China. Mei-Yuh continued to integrate Cortana language understanding models into Microsoft cognitive services. She also shipped a language-model adaptation tool on the same AI platform for customized speech recognition.

Google Scholar

Professional Services

Invited Talks

Awards

Mei-Yuh's personal home page