Ranking biomedical literature search result based on relevance feedback using fuzzy logic and Unified Medical Language System

Online databases and search engines usually return a (long) list of hits that satisfy the user´s search criteria. The returned list of hits is often too long for the user to review every hit if he/she does not know exactly what he/she wants and/or lacks time. Our focus is on biomedical literature search - a healthcare provider needs to find important articles while a patient is waiting for the provider´s diagnosis or treatment decision. In this paper, we developed a fuzzy logic-based ranking approach for biomedical literature search using relevance feedback with the help of Unified Medical Language System (UMLS). UMLS is a biomedical term database that classifies and defines the biomedical language. Relevance feedback refers to an interactive process that helps to improve the retrieval efficiency via user feedback. UMLS provides meaning and semantic type methods that can be used for search result ranking, but they sometimes do not rank the search result accurately. To preliminarily evaluate our proposed approach, we created a document set containing 10 biomedical papers and 20 synthesized documents from them. We designed experiments to: 1) compare the performance of fuzzy ranking method with UMLS meaning and semantic type methods, and 2) evaluate the effectiveness of using relevance feedback in the search process. Our experiments showed that 1) the fuzzy ranking approach improved the average ranking order accuracy by 3.35% and 29.55% as compared with UMLS meaning and semantic type methods respectively, and 2) better ranking result using relevance feedback in the search process.