Text Mining untuk Akuisisi Pengetahuan Secara Otomatis pada Sistem Pakar

Abstract

Automatic knowledge acquisition method from documented source of expert system, now is still residing at under done development stage. This research developed a new method to do automatic knowledge acquisition from text of expert system by combines text mining and neural network technology. The design of research consists of three steps, that is: development of the method, development of the prototype and accuracy level evaluation of the produced knowledge base. Automatic knowledge acquisition method from documented source, which developed at this research consist of six steps, that is: document parsing, topics retrieval, topics clustering, domain extraction, item extraction and solution extraction. Document parsing applied with parsing method of natural language processing to do POS tagging on text. Topics retrieval developed based on information retrieval technology to find studied topics/phrase of text. Topics clustering developed based on self organizing maps technology to group relevant and non relevant topics/phrase. Domain extraction, item extraction and solution extraction developed based on information extraction technology and knowledge graph theory. Domain extraction extracted information of a relevant topic/phrase on text. Item extraction extracted indication information of a domain. Solution extraction extracted solutions information of a domain. Produced knowledge base will represent with object oriented. Inference engine developed based on dempster-shafer theory and fabric fault advisory expert system journal as reference. Automatic knowledge acquisition system prototype and expert system prototype developed with object oriented software engineering approach. Evaluation result show that the produced knowledge base accuracy level is reach 0.6563 and accuracy level can be increase by adding number of knowledge source document.