earticle

영어

In this paper, multi-intelligent agent architecture has been proposed for automatic knowledge extraction from its resources (domain experts and text documents). The extracted knowledge should be stored in a knowledge base to be used later by knowledge-based systems. This article aims to produce an effective knowledge base by cooperation between expert mining and text mining techniques. Firstly, we are constructing an Expert Mining Intelligent Agent (EMIA) able to interview with domain experts for mining problem solving knowledge as production rules in a specific diagnosis domain. It is also responsible for extracting the patterns or linguistic expressions and save it in a conceptual database. Secondly, we are constructing a Text Mining Intelligent Agent (TMIA) capable of extracting production rules from a text document corpus. The achievement of that extraction can be performed by a text document categorization based on a traditional term weighting scheme (TF-IDF) and using the Stanford parser to analyze and produce a parsing tree for each sentence in that document. Then, the TMIA looks for all causal words and takes them as separation words to generate patterns and sub-patterns based on the conceptual database. Finally, the TMIA stores those patterns and sub-patterns in a pre-formatted template and displays it to a domain expert for a modification process to construct accurate production rule.