Consumers and patients are confronted with a plethora of health and health care information, especially through the proliferation of web content resources. Democratization of the web is an important milestone for patients and consumers since it helps to empower them, make them better advocates on their own behalf and foster better, more informed decisions about their health. Nonetheless, they have difficulties in identifying and accessing information that answers their specific questions, through standard information retrieval (IR) and text-based search techniques. This is partly due to vocabulary mismatches between lay terminology and the concepts that underlie medical/technical content. In addition, mental models of patients may not match the conceptual models of content developers, or of content indexers. Although a rich array of tools and vocabularies exists, current term mapping methods and tools for medical professionals are not sufficient to meet patients' needs. The goal of this project is to develop an improved means for patients to search for information resources relating to questions they may have about their health. We will define the characteristics of patient terminology in the context of carrying out web-based IR tasks, and develop a new scalable and flexible patient term mapping and linking method by combining knowledge-based and data-driven approaches. The proposed method uses a semantic network with weighted relations among concepts where the weights signify the semantic closeness between the concepts. The weights and relations can be updated based on prior successful retrievals and can be adapted to specific application contexts. A patient-oriented medical vocabulary tool will be developed that will assist patients in formulating queries for retrieving resources from the Internet. The project will build on existing tools and vocabularies for health information indexing and retrieval, and enhance these resources by incorporating tools and methods that map patient-oriented concepts and mental models to them. The method and the tools developed will be evaluated for their impact on patient IR in a randomized controlled study. Subjects in the study will attempt to retrieve healthcare information pertaining to specific scenarios that they will be provided. We will measure IR precision and recall, discriminating ability, self-perceived success (by subjects), and user-satisfaction with and without the vocabulary support.
Showing the most recent 10 out of 29 publications
