Catherine Lai
I am a post-doc research fellow at the Centre for Speech Technology Research at the University of Edinburgh. My research focuses on how we can use the non-lexical aspects of speech (e.g. speech prosody) to get at what speaker actually mean. I'm currently working mainly on topical and affective information extraction from conversational speech, with a goal of improving assistive technologies. I'm also currently working more generally on developing models of prosody in dialogue with Johanna Moore. We're particularly interested in how we can use idea from topic modeling to get a more robust idea of the relationship between prosody, discourse structure, and information structure. Hopefully this will build some bridges between more theoretical and empirical approaches understanding to this complicated aspect of spoken communication. I recently co-supervised Leimin Tian, who worked on emotion recognition in dialogue. I'm generally interested in how we can characterize aspects of spoken dialogue in automatically derivable ways. The motivation for this is to develop unsupervised approaches for these tasks to enhance automatic dialogue systems. I'm also interested in how we can use speech technology to further social science research and am currently a co-organizer of an interdisciplinary network on Speech, Image and Social Media Data for the Social Sciences
I previously worked on the EU FP project InEvent, which looked at how automatic processing of audiovisual data could be used to aid browsing of large video archives. I mainly worked on how to features often associated with speaker affect, e.g. prosody and measures of participation, can be used for summarization and affect detection. I also ended up doing a bit of HCI evaluation.
Before this, I was a graduate student in the Department of Linguistics at the University of Pennsylvania. At Penn, I worked in the phonetics lab where I took advice from Jiahong Yuan, Mark Liberman, Florian Schwarz, and many other people. My dissertation was about where intonational features, particularly final pitch rises, fit in with semantic and pragmatic theories. I also worked on various other topics ranging from iterated learning in language change, gradability and modality in semantics, tone and stress in Chinese, and prosody of second language learners.
I'm currently part of the UKSpeech organizing committee. In the past, I have served on the Student Advisory Committee of the International Speech Communication Association (ISCA-SAC). I helped organize a workshop on New Tools and Methods for Very-Large-Scale Phonetics. I have also served of the organizing committee of the 2010 Young Researcher's Roundtable on Spoken Dialogue Systems (YRRSDS).
Quite a long time ago, I did a research masters at the University of Melbourne, Australia. I was part of the Language Technology Group and my advisor was Steven Bird. Back in the day, I researched querying and manipulating linguistically annotated structured data.
Research Interests
Prosody, discourse and dialogue structure Affect detection Multimodal language processing Automatic media monitoring