Building an Authoring Tool for Concept-Feature Relationships by Decomposing Phrases in the Text

摘要

We are building a medical student patient simulation system to teach basic science combined with clinical problems in a problem based learning environment. To support the patient simulation system we are building the Intelligent Information Retrieval System (IIRS), which provides five categories of learning resources. A key important part of the system is a customizable thesaurus that depends on the feature-concept relationships. This paper discusses method for building these relationships using the index of a basic science textbook and a clinical text. We claim that clinical science and basic science terms that appear in the same section of Harrison's Principles of Internal Medicine are candidates for inclusion in the thesaurus. We developed rules for phrase construction using the index of a physiology textbook to generate a list of basic science terms. We can provide some possible relevant terms, both basic and clinical, for a particular topic by searching our term databases. In addition, we extract embedded terms in the text, which can not be verified by exact match using our program to refine terms. As a result, we can provide larger range of relevant terms to construct more accurate and detailed relationships.