An architecture for personalized health information retrieval

Abstract

With the rapid proliferation of the Internet, traditional Information Retrieval (IR) techniques need to address challenges that stem from information overload by filtering web documents and ranking them in an order that can be perceived to be more relevant and credible to the end-user. In the domain of care, an increasing number of people turn to the Internet for their and wellness concerns. The results returned by traditional search engines can therefore be overwhelming and, even worse, inaccurate. As a consequence there is a need to design more intelligent web services that pre-process and alter information on the user's behalf. Specifically, this paper describes the design of a personalized search engine that utilizes patient data (either stored in user-managed personal records or in provider-managed electronic medical records) and couples this with a selective crawling of credible medical information to eliminate search results that appear irrelevant to the user (given the user's health profile) and rank the remaining results in order of relevance based on the conditions of users performing the searches. Toward this end, a new ranking algorithm that combines a user's search query and the user's profile is introduced. Finally, comparisons of the search results for users with different profiles and diverse queries are presented using this architecture.