I have been incredibly fortunate to have Professor Michael I. Mandel and Professor Andrew Rosenberg as Ph.D. advisors. During the pursuit of the degree, I am constantly amazed by their wealth of proficient knowledge and remarkable engineering skills.
I studied spoken term detection and speaker state recognition with Professor Rosenberg. We explored different rescoring strategies for the low-resource languages in the BABEL Program, leveraging the signals that are useful in keyword identification but are not being used by the speech recognizer. This research was in collaboration with Columbia University and IBM, providing me a platform to learn from more researchers.
Beyond the BABEL Program, we investigated a variety of acoustic-prosodic features to identify the speaker states, such as emotion, deception, or rating the degree to which a person suffers from the Parkinson’s disease, etc.
Advised by Professor Mandel, I studied automatic speech recognition for noisy environments, and the adaptation approaches for language models. Both the research topics are of immediate significances: most of the traffic that a voice-powered service deals with comes with various noises; most of the data to train a language model in the automatic speech recognition system is written data, whereas the aim is to make predictions on spoken data. For the first problem, we built a system that jointly alleviates the background noise as well as reduces
the speech recognition errors. For the second problem, we explored how to adapt deep neural network based language model from written domain to spoken domain. During the exploration, I received a great deal of guidance and help from my advisor, who is always willing to discuss the research problems as well as the engineering details. I highly appreciate his amazing patience and have benefited a lot from our discussions.
My advisors gave me much freedom to explore a range of research topics interested me. They encouraged and supported me to seek research internships in industry. I have served as an intern at companies ranging from Educational Testing Service, IBM, Nuance to Google. Although not all the projects I have worked for were included in my Ph.D. dissertation, the colorful experiences are a huge treasure.
I also genuinely appreciate the enormous help received from Professor Robert Haralick, who has been such a respectful chair of the computer science department of CUNY-Graduate Center. Life has its ups and downs, when I was down, Professor Haralick showed great understanding as well as offered me timely help. I am extremely thankful for his trust. As a distinguished computer scientist, Professor Haralick also influences me profoundly with his inexhaustible enthusiasm for research and endless brilliant ideas for a multitude of research areas.
After graduating from CUNY, I join Google to work as a software engineer in speech group. I am currently working on automatic speech recognition technology, aiming to help a larger population who speaks various languages embrace the benefits brought by voice search technology. The time at CUNY is a cherished memory and will keep motivating me to make progress in the future.