Advancing the state-of-the-art in human-computer conversation, through work on dialog management, dialog state tracking, user modeling, language understanding, and language generation.
The Conversational Systems Group is part of the Deep Learning Group in Redmond. It is focused on advancing the state-of-the-art in human-computer conversation. Our research spans dialog management, dialog state tracking, user modeling, language understanding, and language generation. We are particularly interested in exploring methods for conversational systems that learn in the course of a dialog.
Building computers which can converse has been a fundamental pursuit of artificial intelligence for decades. Today, advances in modern machine-learning methods provide unprecedented new opportunities for progress. We are tackling challenges such as how to learn a dialog controller for a specific task from very limited interaction; how to create models that combine structured knowledge with machine learning for dialog control; and how a dialog controller can autonomously improve through interaction. We collaborate closely with other groups in Microsoft Research, including Natural Language Processing, Machine Teaching, and Adaptive Systems and Interaction.