Mei-Yuh Hwang

Also known as 黄美玉

Mei-Yuh Hwang is a speech recognition researcher whose work runs from Carnegie Mellon's SPHINX-II recognizer to Microsoft's dictation software, Bing Translator and the Chinese Cortana. At Carnegie Mellon, she named the shared Markov states known as senones in 1992 and completed her PhD under Raj Reddy in 1993. After joining Microsoft in 1994, she led Mandarin speech recognition for the DARPA EARS and GALE projects at the University of Washington from 2004 to 2008. Mobvoi hired her as vice president of engineering in 2016, the IEEE named her a Fellow in 2019, and she worked on search at Microsoft and on scene-text OCR at Meta before retiring in June 2023. Today she is an affiliate professor of electrical and computer engineering at the University of Washington.

Now at: University of Washington
Works: 74
Citations: 2,358
h-index: 24
Badges: Bridge
Boomerang
Loyal

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

1986–1993

SPHINX-II and senones at CMU

She earns a bachelor's degree in computer science at National Taiwan University in 1986 and works on Carnegie Mellon's SPHINX-II speech recognition system from 1987. There she proposes decision-tree clustering of Markov states for continuous speech recognition and names the shared states senones in 1992. In December 1993, she completes her PhD under Raj Reddy with a dissertation on subphonetic acoustical modeling for speaker-independent continuous speech recognition.
1994–2004

Speech products at Microsoft

In 1994, she joins her co-advisor Xuedong Huang on Microsoft's speech recognition research team. Her projects through 2004 include the Whisper dictation system, the Microsoft Speech API, English, Mandarin and Japanese dictation in Office XP, and Microsoft Speech Server. Along the way she ports SPHINX-II speech recognition to Windows, Office and the Speech Server SDK.
2004–2008

Mandarin recognition for DARPA GALE

From 2004 to 2008, she is a researcher at the University of Washington's Signal, Speech and Language Interpretation Lab, where she leads Mandarin speech recognition for the DARPA EARS and GALE projects. Her system's 9.1% character error rate is the best Mandarin recognition result in the official DARPA GALE evaluation of 2007.
2008–2015

Bing Translator and Chinese Cortana

Back at Microsoft Research in 2008, she moves into machine translation, and her team delivers Bing Translator and expands it to more than 40 languages. As a principal science manager at Microsoft China from 2012 to 2015, she leads spoken language understanding for Cortana with teams in Beijing and Suzhou. The Chinese Cortana, Xiao Na, ships on Windows Phone 8.1 in 2014.
2016–2020

Mobvoi's AI lab in Redmond

Mobvoi, a Beijing-based AI company, hires her as vice president of engineering in 2016. She directs the Mobvoi AI Lab in Redmond, Washington, until 2020, with a research agenda of noise-robust speech recognition, hot-word triggering, semantic understanding and multi-turn dialog management.
2019

Elected an IEEE Fellow

The IEEE elects her a Fellow in the 2019 class for contributions to speech and language technology.
April 2020–May 2023

Microsoft Search and Meta

She rejoins Microsoft in April 2020 as a partner science manager for Microsoft Search in Office 365, applying large pretrained language models to Outlook and Teams search and prompting GPT-3 to manipulate Excel spreadsheets in natural language. From May 2022 to May 2023, she is an AI research scientist at Meta in Bellevue, Washington. In 2023, she coauthors DISGO, a paper that proposes word error rates for evaluating scene-text OCR.
As of October 2026

Affiliate professor at UW

From March 2022 to November 2023, she is a remote technical advisor to Vinbrain, and she retires in June 2023. She is an affiliate professor of electrical and computer engineering at the University of Washington.