[Talk] End-to-End Modeling for Abstractive Speech Summarization, Dr Roshan Sharma, Google USA, November 8 (today), 2-3 pm

TITLE: End-to-End Modeling for Abstractive Speech Summarization

TIME AND VENUE: MMCR, EE, C241, 2:00-3:00 pm

In our increasingly interconnected world, where speech remains the most intuitive and natural form of communication, spoken language processing systems face a crucial challenge: they must do more than just categorize speech, they need to truly understand it to generate meaningful responses. One key aspect of this understanding is speech summarization, where a system condenses the important information from spoken input into a concise summary.

In this talk, I will discuss our work on end-to-end modeling for abstractive speech summarization, and expound on our work in long-context modeling, multi-stage training, open source datasets and benchmarks, and finally studies about the impact of various factors on human annotations.

Roshan Sharma is a Research Scientist with Google in New York, USA. He earned his Ph.D. in March 2024 from Carnegie Mellon University, USA for his thesis titled “End-to-End Modeling for Abstractive Speech Summarization”. He has diverse experiences across multiple areas of speech and language processing, including speech recognition, spoken language understanding, noise suppression, multimodal machine learning, and more recently in large-scale foundation models.