Partha Pratim Talukdar ( http://talukdar.net/ ) is a postdoctoral fellow in the Machine Learning department at CMU working with Tom Mitchell on various projects. He is going to talk about one component of Tom Mitchell’s Never Ending Language Learning (NELL) project at CMU. ( http://rtw.ml.cmu.edu/rtw/ )
He will be visiting SFU CS on Tue, May 28 and will give a talk (details below) at 2:30pm in TASC1 9408.
Title: It’s about Time: Temporal Scoping of Relational Facts
Speaker: Partha Pratim Talukdar (CMU)
Recent research has made signicant advances in automatically constructing knowledge bases by extracting relational facts (e.g., Bill Clinton-presidentOf-US) from large text corpora. Temporally scoping such relational facts in the knowledge base (i.e., determining that Bill Clinton-presidentOf-US is true only during the period 1993 - 2001) is an important, but relatively unexplored problem. In this talk, I shall present CoTS (Coupled Temporal Scoping), a joint inference framework for this task. CoTS exploits temporal constraints among facts from within and across relations, and weak supervision in the form of a few labeled examples. In the second part of the talk, I shall present a way to automatically acquire such temporal constraints among relations that CoTS exploits (e.g., actedIn(person, film) typically occurs before wonPrize(lm, award)), given only a database of known facts without time information, and a large document collection. I shall present experimental evidence demonstrating effectiveness of the proposed approaches.