| Current research in biomedicine produces an increasingly large amount of information, and as a result, the manual curation of the literature becomes increasingly difficult. Automatic information retrieval systems are built to help researchers locate relevant information by converting unstructured text into structured, searchable data. This thesis aims to improve the quality of biomedical information retrieval by first developing a method to quickly and accurately identify individual genes in a research article in a process called gene normalization (GN) (Chapter 2). Chapter 3 explores the effects of pooling different GN systems into one meta-system and discusses our contribution to such a system. Thus far, text has played a leading role in biomedical information retrieval, but an overlooked source of valuable information is the mining of article figures. Chapter 4 introduces our approach to improving biomedical image retrieval by using our gene normalization approach with an image search engine that can extract information from within an image. |
