Voices of the Deep: Unsupervised Audio Representations with Masato Hagiwara

On February 23, 2021, machine learning researcher Dr. Masato Hagiwara shares how neural networks can transform speech audio into semantic representations without supervision. Unsupervised audio representation is a fundamental building block for processing, analyzing, and translating human speech and Dr. Hagiwara explains how his experimental results demonstrate how to recover the semantic relationships between words without access to supervision or written texts.
Masato Hagiwara (masatohagiwara.net/) is a natural language processing engineer and researcher at Octanove Labs (www.octanove.com/). He received his Ph.D. degree in Information Science from Nagoya University in 2009. During his Ph.D., he worked at Google and Microsoft Research as an intern and thereafter at Baidu Japan and Rakuten Institute of Technology, focusing on search engine-related language processing research. Most recently, he was working as a Senior Machine Learning Engineer at Duolingo, focusing on educational applications of NLP. He received several paper awards for his work on knowledge acquisition and transliteration. He is the author of "Real-World Natural Language Processing” from Manning Publications (www.realworldnlpbook.com/)
This presentation was part of a series of semi-regular meetups for Voices of the Deep, a multidisciplinary multi-institution project founded and led by scientists Laurance Doyle (SETI Institute), Brenda McCowan (University California at Davis), Fred Sharpe (Alaska Whale Foundation), Michelle Fournet (Cornell University), and Jim Crutchfield (University California at Davis) and supported by Templeton World Charity.
Voices of the Deep: csc.ucdavis.edu/~chaos/share/v...