26 July 2019 - Mike Lewis: Seminar
Reasoning with Generative Models of Natural Language
I will describe several projects in which latent language is generated to support planning and reasoning, instead of just communication. First, I will discuss planning in negotiations, in which self-interested agents must use dialogue to reach a mutually acceptable compromise. Then I will describe generative question answering models (which model the joint distribution over questions and answers), and show how this approach can scale to both broad coverage question answering, and to highly compositional questions involving many steps of logical reasoning. I will conclude with recent work in which agents decompose complex decision making into separate steps of generating a latent plan in natural language, and then executing it.
Mike Lewis is a research scientist at Facebook AI Research in Seattle, working on connecting language and reasoning. Previously he was a postdoc at the University of Washington (working with Luke Zettlemoyer), developing search algorithms for neural structured prediction. He has a PhD from the University of Edinburgh (advised by Mark Steedman) on combining symbolic and distributed representations of meaning. He received an Outstanding Submission Award at the 2014 ACL Workshop on Semantic Parsing, Best Paper at EMNLP 2016, Best Resource Paper at ACL 2017, and Best Paper Honourable Mention at ACL 2018. His work has been extensively covered in the media, with varying levels of accuracy, everywhere from New Scientist to the front page of The Sun.