A long-standing goal for the field of artificial intelligence is to enable computer understanding of human languages. This talk discusses two language understanding tasks to which I have applied machine learning methods in an effort to hasten our progress towards this goal.

The first core requirement in reaching our goal is the ability to transform individual sentences into a form better suited for computer manipulation. This ability, called semantic parsing, requires several knowledge sources, such as a grammar, lexicon, and parsing mechanism. We build on previous research in automating the construction of parsers using machine learning techniques. The result is a combined system that learns semantic lexicons and semantic parsers from one common set of training examples. We demonstrate the performance of the system on a real world domain of answering database queries, comparing it to a previously developed lexicon learner, demonstrating superior performance with our system.

A second core requirement is the ability to carry out conversations. My focus here has been on making the interaction between the system and user more efficient over time due to adjustments on the part of the system. In the second part of the talk, I discuss an adaptive, conversational, speech interface for decision-support tasks, using destination selection as an example. The system unobtrusively collects user preferences from conversation logs, and uses them to build user models which guide future conversations. I will conclude the talk with a discussion of our current research and a brief overview of graduate studies at the University of Utah.

Host: Rajmohan Rajaraman