UTCS Colloquia/AI - Paul Bennett/Microsoft Research, "Class-Based Contextualized Search", ACES 2.402

Contact Name: 
Jenna Whitney
Mar 28, 2011 1:00pm - 2:00pm

There is a sign-up schedule for this event that can be found at



Type of Talk: UTCS Colloquia/AI

Speaker/Affiliation: Paul B

ennett/Microsoft Research

Talk Audience: UTCS Faculty, Graduate Stude

nts, Undergraduate Students, and Outside Interested Parties


e: Monday, March 28, 2011, 1:00 p.m.

Location: ACES 2.402


: Raymond J. Mooney

Talk Title: Class-Based Contextualized Search

nTalk Abstract:
Information retrieval has made significant progress in re

turning relevant results for a single query. However, much search activity
is conducted within a much richer context of a current task focus, recent
search activities as well as longer-term preferences. For example, our ab

ility to accurately interpret the current query can be informed by knowledg

e of the web pages a searcher was viewing when initiating the search or rec

ent actions of the searcher such as queries issued, results clicked, and

pages viewed. We develop a framework based on classification that enables r

epresentation of a broad variety of context including the searcher''s long-

term interests, recent activity, and current focus as a class intent dist

ribution. We then demonstrate how that can be used to improve the quality o

f search results. In order to make such an approach feasible, we need reas

onably accurate classification into a taxonomy, a method of extracting and
representing a user''s query and context as a distribution over classes,

and a method of using this distribution to improve the retrieval of relevan

t results. We describe recent work to address each of these challenges. Thi

s talk presents joint work with Nam Nguyen, Krysta Svore, Susan Dumais,

and Ryen White.

Speaker Bio:
Paul Bennett is a Researcher in the Con

text, Learning & User Experience for Search (CLUES) group at Microsoft Res

earch where he works on using machine learning technology to improve inform

ation access and retrieval. His recent research has focused on classificati

on-enhanced information retrieval, pairwise preferences, human computatio

n, and text classification while his previous work focused primarily on en

semble methods, active learning, and obtaining reliable probability estim

ates, but also extended to machine translation, recommender systems, and
knowledge bases. He completed his dissertation on combining text classifie

rs using reliability indicators in 2006 at Carnegie Mellon where he was adv

ised by Profs. Jaime Carbonell and John Lafferty.