Modeling of information needs for medical information retrieval.

摘要

As more medical information is made available electronically, there is a growing need for tools that provide efficient access to these collections of digital information. This thesis develops an approach to help physicians recognize and articulate their information needs precisely for the purpose of facilitating online searches. I explore a conceptual model, explicating the semantic patterns that underlie physicians' information needs, and employ the model to anticipate their needs when searching information in the context of clinical care. The model is evaluated and refined in the context of three empirical studies.; I developed a formal representation of an information need that uses a domain-specific ontology and conceptual graph formalisms to express the semantics of physicians' information needs. The semantic patterns of information needs, called generic queries, are used in conjunction with patient data to generate patient-specific clinical questions for suggestion. I developed and implemented an algorithm that determines generic query's degree of applicability to a particular clinical situation.; To enhance an understanding of physicians' information needs, I conducted three empirical studies: (1) a conceptual analysis of a taxonomy of physicians' information needs developed by Ely et al. (2000); (2) an investigation of semantic structures extracted from the medical literature; (3) a conceptual analysis of information needs expressed by physicians when they review three clinical scenarios. I compared and contrasted the semantic structures extracted across the studies. The three studies revealed similar semantic patterns that converge into a relatively small number of semantic structures. This confirms the validity of our observation and strengthens the viability of the approach.