The major objective of the project is the development of methods for automatic processing of natural medical language. Crucial information of patient records is embedded in natural language text generated during physical exams or in reports from various hospital laboratories. Precise retrieval of subsets of patient data via this information is needed to increase the scientific value of this latent information pool for retrospective studies, including studies of drug effects, and for teaching purposes. The spreading of this technology beyond a limited domain depends upon developing """"""""intellegent"""""""" algorithms able to learn the highly particular semantic models and language syntax governing the language used in specific micro-domains of medical knowledge. Developed technology is utilized in the automatic encoding of Surgical Pathology reports for NCI/DCBD/LP, which become a source of problems and examples in medical lexicography and semantic modeling of medical information. A logic-based software platform (Lexicographic Environment Software) is under development to facilitate manipulation of acyclic directed graph dictionaries and research in medical lexicography, """"""""high resolution"""""""" representation and query of medical information, and induction and evolution of syntactic and semantic rules. Experiments with artificial neural network solutions to problems arising in natural language were performed.
