Toyota AI Seminar – Amanda Stent

In this talk I will first present desiderata for functions of a dialog manager in today's commercial and academic environments (including course advising systems!), and then briefly survey dialog manager technologies over time, from finite-state models to the latest statistical methods. I will highlight gaps between real-world dialog needs and existing dialog management techniques. I will describe a new statistical approach, Task Lineage-based Dialog State Tracking (TL-DST), aimed at seamlessly orchestrating multiple tasks with complex goals across multiple domains in continuous interactions, and present evaluation results for this approach.
Amanda Stent is an NLP researcher at Bloomberg in New York. She has previously worked as a director of research at Yahoo, a principal member of technical staff at AT&T Research, and an associate professor at Stony Brook University. Her areas of interest include discourse, dialog and natural language generation. She is the author of over 75 peer-reviewed publications and several patents, is president of the ACL/ISCA Special Interest Group on Discourse and Dialogue and on the editorial board of the open-access journal Dialogue & Discourse, and serves on the board of CRA-W.