Research goal: to make archives of webcast lectures & presentations the equivalent of digital libraries.

Focus: improve the usefulness and usability of automatically-generated transcripts of webcast lectures and presentations.

Research themes:

• measuring user needs for transcription accuracy
• engaging users to collaborate (wiki) on editing imperfect ASR transcripts
• better transcripts for lectures by using external knowledge sources (WWW, lecture slides)
• language modelling for the conversation- and the topic-specific styles of lectures
• lecturer-specific conversational modelling by exploiting the user corrections of transcripts
• development of a user-motivated measure of transcript quality

Approach: combine research in Human-Computer Interaction and Automatic Speech Recognition that would ultimately see text transcripts of lectures and presentations being integrated in webcast archives.

Download: a printable overview of this research and an abstract of my thesis proposal.