Knowledgeintensive and statistical approaches to the retrieval and

[摘要] Retrieving and annotating relevant informa tion sources in the genomics literature are dif ficult but common tasks undertaken by biolo gists. The research presented here addressesthese issues by exploring methods for retriev ing MEDLINE citations that answer real bi ologists’ information needs and by addressingthe initial tasks required to annotate MED LINE citations having genomic content withterms from the Gene Ontology (GO). We ap proached the retrieval task using two methods:aggressive, knowledgeintensive query expan sion and text neighboring. Our approaches tothe triage subtask for annotation consisted oftraditional machine learning (ML) methods aswell as a novel ML algorithm for thematicanalysis. Finally, we used a statistical, ngramheuristic to decide which of the GO hierar chies should be used to annotate a givenMEDLINE citation.

[学科分类] 社会科学、人文和艺术（综合）