Natural Language Inference in Natural Language Terms
sala d'actes de la FIB - Edifici B6 (confirmat)
4 Setembre 2012
12:00h - Presentació
Reasoning requires a framework for meaning representation, over which inferences would be applied. Traditionally, logic was perceived as the mathematical vehicle for inference, and a plethora of logic languages were invented as meaning representations. As for inference over textual expressions, the classic NLP approach followed the logic-based paradigm: first translate natural language into logic and then conduct logical inference. However, a more feasible alternative evolved in common practice: apply inferences directly over linguistic representations, such as parse trees and lexical co-reference chains, even though these were originally invented to capture language structure rather than as meaning representations for inference. In this talk I will describe recent learning-based approaches that perform broad textual inference over common linguistic representations, under the textual entailment paradigm. I will suggest that we should invest substantially in developing generic inference mechanisms over human-language representations, while possibly interfacing with "extra-linguistic" inferences where needed. Time permitting, a short demo of the open-source BIUTEE inference system will be presented.
Ido Dagan holds B.Sc. (Summa Cum Laude) and Ph.D. degrees in Computer Science from the Technion, Israel. He conducted his Ph.D. research in collaboration with the IBM Haifa Scientific Center, where he was a research fellow in 1991. During 1992-1994 he was a Member of Technical Staff at AT&T Bell Laboratories. During 1994-1998 he has been at the Department of Computer Science of Bar Ilan University, to which he returned in 2003. During 1998-2003 he was co-founder and CTO of a text categorization startup company, FocusEngine, and VP of Technology at LingoMotors, a Cambridge Massachusetts company which acquired FocusEngine.