Prof. Philip Cohen - Back to the Future for Dialogue Research

Organized by: 
James L. Crowley,
Prof. Philip Cohen - Laboratory for Dialogue Research, Monash University


Dr. Philip Cohen has long been engaged in Artificial Intelligence, specifically in the subfields of human-computer dialogue, multiagent systems, and multimodal interaction. He is a Fellow of the Association for the Advancement of Artificial Intelligence, and a past President of the Association for Computational Linguistics. Currently he is Professor of Artificial Intelligence, and Director of the Laboratory for Dialogue Research, Faculty of Information Technology, Monash University. He has also held positions as Chief Scientist, AI and Sr. Vice President for Advanced Technology at VoiceBox Technologies, at Adapx Inc. (founder), the Oregon Graduate Institute, Artificial Intelligence Center of SRI International, Fairchild Laboratory for Artificial Intelligence, and Bolt Bernanek and Newman.

His accomplishments include co-developing influential theories of intention, collaboration, and speech acts, pioneering research in the field of multimodal interaction, co-developing and deploying multimodal systems for the US Government, and conceiving and leading (through 1994) the project at SRI International that developed the Open Agent Architecture, which eventually became SiriTM. At Voicebox, he led the research and development of next generation conversational agent technologies.

Cohen has published more than 150 refereed papers, with more than 17,000 citations (per Google Scholar , and received 5 patents. His paper with Prof. Hector Levesque “Intention is Choice with Commitment” was awarded the Inaugural Influential Paper Award from the International Foundation for Autonomous Agents and Multi-Agent Systems in 2006. In 2017 he was the recipient of the Sustained Accomplishment Award from the International Conference on Multimodal Interaction. He is co-author of the book The Paradigm Shift to Multimodality in Contemporary Computer Interfaces, ACM Press (2015), and a co- editor of the 3-volume Handbook of Multimodal-Multisensor Interfaces, ACM Press (2017-2019).

This talk argues that future conversational systems that can engage in multi-party, collaborative dialogues will require a more fundamental approach than existing “intent + slot”-based systems. I identify significant limitations of the state of the art, and argue that returning to the plan-based approach todialoguewillprovideastrongerfoundation. Finally,Isuggestaresearchstrategythatcouplesneural network-based semantic parsing with plan-based reasoning in order to build a collaborative dialogue manager.