Mei-Yuh Hwang 黄美玉

Mei-Yuh received her PhD in Computer Science from Carnegie Mellon University in 1993 and have worked in AI industry/academia for three decades, publishing numerous conference and journal papers, and delivering industry products in speech recognition, machine translation, language understanding, and image representation & recognition. Mei-Yuh's focus has been always on turning state-of-the-art technologies into end-users' hands. She is an IEEE fellow, who is passionate in bridging the gap between academia and industry.

Vinbrain, 3/2022-11/2023

• Technical advisor, remote
• Attendee, RSNA 2023

Meta, 5/2022--5/2023

• AI Research Scientist, Bellevue WA
• Speech recognition patent: Device wake-up modeling using a word-piece model.
• Scene-Text OCR
  • DISGO: Automatic End-to-End Evaluation for Scene Text OCR: Evaluating machine translation from scene-text OCR output
• Text-to-image generation

Microsoft, 4/2020-4/2022

• Partner science manager, Microsoft Search for Office 365, Bellevue WA.
• NLP for Outlook and Teams search, based on fine-tuning on, and/or few-shot prompting to, various large pre-trained language models.
• CTO winning project: Prompt engineering on GPT-3 copilot to manipulate Excel spreadsheets via human natural languages.
  See related post in GPT-3 application.

Mobvoi AI Lab, 2016-2020

• Director, Redmond WA.
• Mobvoi makes speech-enabled smart IoT devices, from hardware to software, all in-house.
  • We are one of the most successful companies in smart watches in China.
  • Mobvoi also provides the in-car voice assistant for VW automobiles in China. The technology includes both on-the-cloud and on-device voice navigation, enabling majority of services without internet connection.
  • We are active in generative AI and have been very successful in personalized TTS.
  • Though a young company that focuses on industry products, we are actively participating in our speech research community with our limited resources.

Microsoft China, 2012-2015

• Principal science manager, Spoken language understanding for Cortana, Beijing and Suzhou.
• Non-English Cortana
  • 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 semantic slot tags.
  • The protoype model was further improved via iterative data augmentation using RNN and newly logged data.
• Microsoft cognitive services on Azure cloud:
  • Cortana non-English language understanding for LUIS,
  • Language-model adaptation for customized speech recognition.
  • Spontaneous speech recogition for Skype speech-to-speech translation.

Microsoft and UW, 1994-2012

• Machine translation, Microsoft, Redmond WA, 2008-2012.
  • Co-built Bing Translator automated training infrastructure, including the design and implementation of map-reduce parallel processing, based on DryadLink.
  • Designed and implemented Bing Translation Hub for customized vertical-domain translation.
• Speech recognition at Univ. of Washington, Seattle WA, 2004-2008.
  • Led the DARPA EARS and GALE Mandarin speech recongition projects at University of Washington.
  • Won the best Mandarin speech recognition.
• Speech recognition, Microsoft Research & Production, Redmond WA, 1994-2004.
  • Ported Sphinx-II speech recognition to Microsoft on Windows desktop, Office, and Microsoft Speech Server SDK, for the recognition of multiple languages.

SPHINX-II speech recognition at CMU, PA, 1987-1993

• 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) had been widely adopted for two decades since its inception, until recent years when end-to-end neural-transducer based and transformer-based speech recognition took over a new era.
• Participated in numerous DARPA speech recognition evaluation benchmarks (Resource Management, WSJ, ATIS) and won the top position consistently.

Awards

• 2023, Outstanding Reviewer, ICASSP 2023
• 2021, AAIA Fellow
• 2019, IEEE Fellow
• 2010, Microsoft Gold Star Award from Microsoft Research, Redmond, WA
• 1992, Allen Newell Research Excellence Medal, Pittsburgh, Carnegie Mellon University
• 1986, Phi Tao Phi Scholastic Honor Society, recommended by National Taiwan University

Professional Services

• 2015-2018: IEEE ISCSLP steering committee
• 2013: IEEE associate editor for Transactions on Audio, Speech, and Language Processing (ASLP)
• 2011: Technical Chair of IWSLT
• 1998: Publicity Chair IEEE ICASSP
• Reviewers for IEEE Transactions on ASLP
• Technical committee for ICASSP, Interspeech, ISCSLP, ACL, NAACL

Invited Talks

• 2020: WeCNLP Summit, Seattle online.
• 2019: Northwestern Polytechnical University, Xi-an, China
• 2018 April: Yuanchuan Telephony company, Taiwan
• 2017: UWEE Research Colloquium Talk
• 2017: Panelist, CMU Summit
• 2017: Talk at Southwest Forestry Univerisity, Kunming, China
• 2017: Talk at Soochow University, Suzhou, China
• 2015: Talk at Soochow University, Suzhou, China
• 2014: Talk at University of Science and Technology of China, Suzhou campus
• 2014: Keynote speech at PhD Forum, Northwestern Polytechnical University, Xi-an, China
• 1994-2003: Lecturer at National Taiwan University, Academia Sinica, and ITRI, Taiwan
• 1993: Lecturer at IBM Gaithersburg

Google Scholar

Education

• PhD, Computer Science, Carnegie Mellon University, December 1993.
• BA, Computer Science, National Taiwan University (台大), June 1986.