Graduate students, staff, and faculty are cordially invited to attend the SLIS Colloquium Series.
Title: WIDIT in TREC-2003 Web Track
Speaker: Kiduk Yang, School of Library and Information Science
Date: Friday, December 12, 2003
Time: 2:00-3:30pm
Location: LI 001 Main Library
Lecture preceded by an informal gathering with cookies, tea, and coffee, available at 2:15pm.
Abstract:
This talk discusses SLIS's participation in the TREC-2003 Web Track. The Text REtrieval Conference (TREC) supports research in the information retrieval community by providing the infrastructure necessary for large-scale evaluation of text retrieval methodologies since 1992. TREC consists of a set of tracks that focus on particular retrieval tasks (e.g., Filtering, Genome, Question Answering, Video, Web). The Web track features two search tasks on a 18 gigabyte (1.25 million web pages) subset of .GOV website: Topic Distillation (TD) task to find a list of key resources for a particular topic, and Named Page finding (NP) task to find the specific page described by its name (e.g. 'Internal Revenue Service' -> 'www.irs.gov'). SLIS's approach to both tasks involved creating separate indexes of body, anchor, and header texts and combining the results of using different indexes, as well as retrieval performance optimization by result re-ranking based on phrase and acronym matching, link counts and site compression. In addition to the discussion of the TREC Web track research, the presentation will include a brief overview of TREC and a description of WIDIT, which was used to conduct the Web track experiments. For more information about TREC, see: http://es.cmis.csiro.au/TRECWeb/guidelines_2003.html
Bio:
Kiduk Yang is assistant professor of Information Science. His research interests include information retrieval (IR), knowledge organization, web data mining and fusion of IR methods. He has been an active participant in TREC (http://trec.nist.gov/) since 1997 and is the main architect of an experimental retrieval system called IRIS (http://ils.unc.edu/iris/). He is currently developing a Web-based application environment for IR research called WIDIT (http://widit.slis.indiana.edu/index.shtml).
Posted November 14, 2003