The Dirichlet-Multinomial Model for Bayesian Information Retrieval

by I-Li Lu.

Abstract: We investigate a retrieval model using Dirichlet-Multinomial distribution and show that it provides a plausible characterization of the Bayesian retrieval process without the assumption of conditional independence. We provide theoretically justified algorithm for information retrieval using the Bayesian framework. We illustrate that the concepts of minimaxity and admissibility from statistical estimation theory may be used for the selection of initial estimates.

Key Words: Relevance feedback, beta-binomial, admissible minimax

Author:
I-Li Lu, i-li.lu@boeing.com

Editor: Roger Peck, rpeck@csub.edu

READING THE ARTICLE: You can read the article in portable document (.pdf) format (349720 bytes.)

NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.

This page has been accessed 2532 times since October 30, 2007.


Return to the InterStat Home Page.