Applying machine learning to small group interactions has the potential to be very useful for facilitating meetings, understanding group dynamics, and providing real-time feedback to group and teams. The field of Social Signal Processing includes much work along these lines, with example tasks including prediction of group performance and detection of emergent leaders. However, that work has focused almost entirely on nonverbal features. There has been a surprisingly small amount of research on using natural language processing for understanding and predicting small group phenomena. This talk will present a variety of tasks and experimental results demonstrating that NLP can be useful for predicting aspects of group interaction, such as predicting group performance on a task, predicting post-task assessments and sentiment from the participants, and learning about unobserved group behaviours through meeting artifacts. It will also be argued that language-based predictive models are very valuable when we need to provide interpretable models or actionable feedback to a group -- two scenarios where nonverbal models on their own may be insufficient.
Gabriel Murray is an Associate Professor in Computer Information Systems at University of the Fraser Valley, and an Affiliate Professor in Computer Science at University of British Columbia (Canada). His research primarily focuses on the intersection of speech and language processing and small group interaction. He teaches a variety of courses related to artificial intelligence, including machine learning and natural language processing. He received his PhD in Informatics from the University of Edinburgh, under the supervision of Drs. Steve Renals and Johanna Moore.
Host: Oliver Lemon