Proc AMIA Symp. 1999 ; (1-2): 1210.

Improved Access to Alternative Medicine.

Liddy ED.

Syracuse University, Syracuse, NY.

At present, the available medical literature is searchable primarily with Boolean search engines using difficult-to-learn controlled vocabulary. These two facts make easy access to medical information by healthcare providers and patients unnecessarily difficult. To improve on this situation, we have been working to create MedLink, a state of the art Natural Language Processing system, which can provide efficient and effective access to alternative/ complimentary medical literature. MedLink, which is being funded by a Phase I SBIR grant from NIH, has a dual goal for Phase I: 1) to develop a finely-tuned, immediately usable information system for alternative medicine, and; 2) to determine, on an experimental level, what levels of language are critical to optimizing retrieval within the medical subject domain. Phase I has resulted in the development of a prototype MedLink system, which is based on TextWise's DR-LINK Retrieval system, which in turn is based on a full linguistic processing model which simulates the multi-level language understanding of humans. By interpreting words, the syntax, the semantics, and the discourse structure of the written document itself, DR-LINK is able to retrieve information with precision and understanding. While DR-LINK is operational, we have spent the past few months attempting to answer the question of which are the key elements of the medical sublanguage to be adapted in order to optimize information retrieval for a medical domain. The results of this research will assist the MedLink system developers in Phase II to create an optimized system for integrated medical information access based on a solid understanding of the impacts of each level of linguistic understanding.