Events

James Fan: “Interpreting Loosely Encoded Knowledge”

April 27, 2006 | USC ICT

Speaker: James Fan
Host: Jonathan Gratch

Knowledge based systems are capable of answering questions that have elaborate contexts and require deep reasonings. Unlike information retrieval based question answering, knowledge based systems take in complex encodings from users and return the responses by applying reasoning mechanisms on symbolic knowledge bases. A key challenge in using knowledge based question answering systems is how to align a user’s question encoding with the idiosyncrasies in the existing knowledge base. We call such misalignments “loose speak”. I will present the results of our study of loose speak, and describe a loose-speak interpreter which automatically aligns an encoding with the an existing knowledge base. More details of the loose-speak interpreter and its evaluations can be found in the publications section of http://www.cs.utexas.edu/~jfan.