MEDQUAL: Improving Medical Web Search over Time with Dynamic Credibility Heuristics

Abstract

Performing a search on the World Wide Web (WWW) and traversing the resulting links is an adventure in which one encounters both credible and incredible web pages. Search engines, such as Google, rely on macroscopic Web topology patterns and even highly ranked 'authoritative' web sites may be a mixture of informed and uninformed opinions. Without credibility heuristics to guide the user in a maze of facts, assertions, and inferences, the Web remains an ineffective knowledge delivery platform. This report presents the design and implementation of a modular extension to the popular Google search engine, MEDQUAL, which provisions both URL and content-based heuristic credibility rules to reorder raw Google rankings in the medical domain. MEDQUAL, a software system written in Java, starts with a bootstrap configuration file which loads in basic heuristics in XML format. It then provides a subscription mechanism so users can join birds of feather specialty groups, for example Pediatrics, in order to load specialized heuristics as well. The platform features a coordination mechanism whereby information seekers can effectively become secondary authors, contributing by consensus vote additional credibility heuristics. MEDQUAL uses standard XML namespace conventions to divide opinion groups so that competing groups can be supported simultaneously. The net effect is a merger of basic and supplied heuristics so that the system continues to adapt and improve itself over time to changing web content, changing opinions, and new opinion groups. The key goal of leveraging the intelligence of a large-scale and diffuse WWW user community is met and we conclude by discussing our plans to develop MEDQUAL further and evaluate it.