Book Cover

The 2002 Spring Symposium Series

Stanford University

Sponsored by the American Association for Artificial Intelligence
Stanford University
Symposium on Mining Answers from Texts and Knowledge Bases

March 25-27, 2002

The rate of producing textual documents is quite larger than the rate of generating reliable knowledge bases and reasoning mechanisms. However, the information expressed in various on-line textual documents, either on the Internet or in large text repositories cannot be computationally used unless it is associated with expert knowledge bases. The incorporation of textual information into knowledge bases is not simple, due to the multiple forms of ambiguities that characterize natural language texts. However, today we are in the position of having sufficiently large knowledge bases available and the natural language processing technologies have matured enough to process real-world documents and extract information and answer natural language questions with good accuracy.

Part of the recent success of open-domain Q/A is due to novel combinations of technology developed in the 90s (e.g. named entity recognizers) with techniques used in the 80s (e.g. abductive interpretations of texts) and novel indexing/retrieval mechanisms (e.g passage retrieval). Discovering relevant knowledge from the web can be combined with domain knowledge provided by large-scale knowledge bases (e.g. Cyc, Wordnet, UT's Component Library, IEEE's Standard Upper Ontology effort). Furthermore, textual information extraction techniques can be enhanced by using world knowledge available in large lexicon-semantic knowledge bases.

Achieving orders of magnitude improvement in question answering performance requires us to make synergistic use of these advances. Extraction and text mining methods must make use of knowledge based inference in support of the extraction task and for post processing the extracted information. Knowledge bases need to rely on large corpuses of knowledge to support their initial creation, and then subsequent testing and maintenance.

The symposium will bring together diverse techniques for text and answer mining from AI (more specifically, natural language processing, machine learning, knowledge representation and reasoning) with information retrieval (from text collections or from the web) or data tracking and detection. The invited talks and contributed papers will focus on topics such as common sense knowledge bases, linguistic knowledge bases, use of dialog in query formulation, inference and query evaluation techniques, and techniques for competence evaluation of question answering systems.

Organizing Committee

Sanda Harabagiu (cochair), University of Texas at Dallas (; Vinay Chaudhri (cochair), SRI International (; Bruce Porter, University of Texas Austin; Ray Mooney, University of Texas Austin; Tom Mitchell(tentative) Carnegie-Mellon University; Claire Cardie, Cornell University; Richard Fikes, Stanford University; Dan Moldovan, University of Texas at Dallas; Srinivas Narayanan, SRI International; Donna Harman, NIST.

Symposium Schedule

Monday March 25th 2002
9:00--9:30 Opening RemarcksSanda Harabagiu
9:30--10:30 Advanced Question Answering - Plenty of challenges to go around See presentation slides. John Prange
10:30--11:00 Break
11:00--11:30 Abductive Processes for Answer Justification Sanda Harabagiu and Steve Maiorano
11:30--12:00Mining Answers for Causation Questions Roxana Girju and Dan Moldovan
12:00--12:30 Processing Definition Questions in an Open-Domain Question Answering System Marius Pasca
12:30--14:00 Lunch Break
14:00--14:30 Text Mining with Information ExtractionUn Yong Nahm and Ray Mooney
14:30--15:00Point and Paste Question AnsweringSanda Harabagiu, Finley Lacatusu and Paul Morarescu
15:00--15:30An Evolutionary Genre-Based and Domain Independent Approach to High Level Knowledge Discovery in TextsJohn Atkinson, Stuart Aiken and Chris Mellish
15:30--16:00 Break
16:00--16:30Gleaning Answers from the Web Nick Kushmerick
16:30--17:00 The Use of Question Types to Match Questions in FAQ Finder Steve Lytien and Noriko Tomuro
17:00--17:30 Automatically Identifying Candidate Treatments from Existing Medical Literature Catherine Blake and Wanda Pratt
17:30--18:00 Break
18:00--19:00 Reception
Tuesday March 26th 2002
9:00--10:00 Panel on Large Knowledge Bases Adam Pease, Tony Cohn, Pat Hayes, Ken Murray and Chris Welty
10:00--10:30 Automatic Deduction for Spatial ReasoningTomas Uribe
10:30--11:00 Break
11:00--12:00 Finding Similar Content in Different Documents John Everette
12:00--12:30Analogical Reasoning on Large Knowledge Bases Shawn Nicholson
12:30--14:00 Lunch Break
14:00--15:00 Peer Data Management Systems: Planning for the Semantic WebAlon Halevy
15:00--15:30 Mining Answers from Texts and KBs: Our PositionBruce Porter, Ken Barker, James Fan, Paul Navratil, Dan Tecuci and Peter Yeh
15:30--16:00 Break
16:00--17:00 Explaining Knowledge Bases for Query Answering on Semantic WebDeborah McGuiness
17:00--17:30 Searching for Narative StructuresHenrik Scharfe
17:30--18:00 Break
18:00--19:00 Plenary Session with other Symposia
Wednesday March 27th 2002
9:00--10:00 CYKCORP's View of Question AnsweringMichael Witbrock
10:00--10:30 AskMSR: Queston Answering Using the Worldwide Web Michele Banko, Eric Brill, Susan Dumais and Jimmy Lin
10:30--11:00 Break
11:00--11:30 Qanda and the Catalyst Architecture Scott Mardis and John Burger
11:30--12:00 Closing RemarcksVinay Chaudhri