The methodology of knowledge acquisition from the collection of IR and UV spectra

An expert system for identification of parts of chemical molecules from IR and UV spectra is described. To develop the system, the intelligent knowledge engineering program environment SCANKEE¹ (elaborated at the Department of Computer Chemistry) was applied. The expert systems for identification of a chemical compound described in the literature use hand-crafted knowledge bases. To eliminate the time-consuming writing of the rules, a module for the automatic creation of rule knowledge bases was developed. The knowledge derives from the collection of molecular spectra of organic compounds and their structural formulae. The algorithm for the automatic generation of the production rules (in the rule knowledge base) is realized in two steps: (i) the creation of the correlation tables and (ii) the generation of the production rules base.