The author has developed a general authoring CAI system, incorporating a reasoning function, called IROSA-II (Information Retrieval Oriented CAI Support system). He used a semantic network, one of the knowledge representation technologies evolved from research on artificial intelligence for its architecture. IROSA-II makes use of an improved semantic net model, one in which property lists for each keyword, i.e., the anticipated answers, are defined for each semantic frame. The property list attributes include superordinate concepts, superordinate sets, superordinate attributes, parts of speech, synonyms, antonyms, applied fields and examples. The property lists for each semantic net frame are integrated into the semantic net dictionary. This can be considered a type of knowledge base incorporating a semantic net structure. The inference engine has functions similar to those found in response string analyzers. The inference program, incorporating a semantic net dictionary search program, develops inclusive relationships among words thereby providing access to the knowledge base. Results of the production and student testing of several experiment courseware units show that IROSA-II may be a very useful tool for teaching the semantic world of facts, concepts and procedures. Developing this AI-oriented CAI system involved changing the paradigm from programmed learning CAI to intelligent CAI, a change which permits, at least within limited conditions, an intellectual dialogue between computer and student. This report, in relation to the development of IROSA-II, covers a model for knowledge representation through use of table-like semantic nets, methods for developing inferences from the semantic net table and the results of evaluation of experimental courseware.
View full abstract