Knowledge-intensive conceptual retrieval of biomedical literature

Biologists search for biomedical literature on a daily basis. A concept-based approach via utilizing biomedical lexical resources was developed to improve the retrieval effectiveness of biomedical literature. A database of abbreviations and their definitions was created from MEDLINE.;A number of techniques within this approach were developed and evaluated, including: (i) mapping query terms to MeSH terms and UMLS concepts; (ii) adding synonyms, hyponyms and hypernyms of query terms, (iii) adding lexical variants of gene symbols and implicitly related terms; (iv) utilizing two similarity functions, one measuring concept similarities between the query and the documents and another one measuring term similarities between the query and documents; (v) disambiguation of abbreviations; and (vi) utilizing abbreviated query terms to match related terms in documents.;This concept-based approach was applied to two tasks that address two important information needs of biologists. TASK I: Answering questions of biologists regarding the relationships between biological objects and biological processes; TASK II: Finding biological entities that satisfy certain conditions from texts.;The findings reveal that incorporating domain-specific knowledge by utilizing biomedical lexical resources in a concept-base approach significantly improved the retrieval effectiveness of our information retrieval system and it is critical to the success of such a biomedical IR system. The major challenge in our future work is how to utilize these biomedical domain knowledge (or resources) more appropriately.