Improving the performance of retrieval systems remains a major goal of information science. With millions of biomedical citations available online, a practical approach is the development and implementation of add-on devices to enhance their performance instead of a total retooling of our existing system. Analysis shows that each retrieval transaction depends on the success with which (1) the query is interpreted, (2) the document file is represented, and (3) the strategy is utilized to interrogate the file. Clearly, the system's ability to establish the relevance of documents to any given query holds the key to the system's efficacy. Unfortunately, the notion of relevance has not been formally defined, nor precisely measured. In all operating systems, two approaches have been used independently of each other, namely, retrieval by descriptors (semantic relevance) and by citations (pragmatic relevance). Evidence shows a lack of understanding of the difference in retrieval outcomes of each. A serious research gap exists in the study of the combined use of these two modes of relevance. We propose to construct a test database, the content of which is represented by descriptors and by citations, so that experiments may be conducted to analyze and compare the outcome characteristics by both types of retrieval. Furthermore, independent of document representations, a number of alternative search strategies will be tested for retrieval effectiveness. Algorithms incorporating semantic and pragmatic retrieval capabilities may be developed. Add-on devices implementing such enhancements represent a realistic and effective approach to improving our online systems.
| Pao, M L (1993) Perusing the literature via citation links. Comput Biomed Res 26:143-56 |
