Events Calendar

CS Colloq: Dr. Kenji Sagae

Talk Title: Practical analysis of natural language syntax, semantic roles and discourse structure with shift-reduce algorithmsSpeaker: Dr. Kenji SagaeHost: Prof. Kevin KnightABSTRACTAutomatic analysis of the structure of natural language through syntactic parsing techniques has long been considered of great potential value in the study of language, the development of language-enabled systems and interfaces, and the application of language technologies (such as machine translation, question answering and text mining) to the rapidly growing body of information in the form of machine readable text. However, for many years parsing systems suffered from lack of robustness and efficiency to deal with large-scale tasks. Recent research on linear-time parsers that learn from annotated data has opened new possibilities for how these and other issues in practical parsing technologies can be addressed.In this talk I will first present a simple and effective parsing framework that addresses the main challenges in the deployment of parsing technologies in practical tasks. I will show how the combination of machine learning and a parsing approach inspired by Knuth's deterministic LR algorithm produces parsers that are fast, robust and accurate. I will also present extensions of this framework that allow for linear-time analysis of semantic roles and discourse structure, and discuss the application of the resulting data-driven shift-reduce parsing approach in areas as diverse as child language analysis, biomedical text mining, and virtual human dialogue systems.BIOKenji Sagae is a research scientist in the Institute for Creative Technologies at the University of Southern California, where he works on natural language processing for virtual humans and related systems.
Before joining ICT as a research associate in June of 2008, he was a research associate in the Computer Science department of the University of Tokyo, where he worked on the connection of data-driven parsers to theoretically-motivated syntactic models, and the application of natural language processing to information extraction in bioinformatics. He received a PhD in Language Technologies from the School of Computer Science at Carnegie Mellon University in 2006. His dissertation research focused on automatic syntactic analysis of transcripts of dialogues between children and adult caregivers. He is currently the Information Officer for SIGPARSE, the international interest group on parsing technologies, and his parsing software is used by several research groups in the areas of child language and information extraction for biomedical text.

Location: Seaver Science Library (SSL) - 150

Audiences: Everyone Is Invited

Contact: CS Front Desk