Establishing and Retrieving Domain Knowledge from Semi-Structural Corpora

1.1 Knowledge representation The most essential part of building an expert system is the acquirement and representation of domain knowledge. In the seventies, Feigenbaum indicated the important concept of knowledge engineering. He emphasized that to utilize knowledge in problem-solving process is equally important with knowing how to solve a problem. Knowledge, according to how it is stored, can be classified to tacit knowledge and explicit knowledge. The tacit knowledge, existing in the brain of experts, can only be acquired through interviewing the domain experts. On the other hand, the explicit knowledge can be expressed clearly. Since explicit knowledge is easier to be handled, it was used in most expert systems. Knowledge representation affects how problems are solved. Human knowledge can be expressed in the form of mathematic formulas, speech, text and figures. In artificial intelligence domain, especially in expert system research, several knowledge representation forms had been proposed (Negnevitsky, 2002). They are: 1. Semantic networks (Quillian, 1965, 1968): Using directed graph to represent knwolege objects and their relationship. Each object in the network is linked to other objects by their semantic relationships. 2. Case-based format (Watson, 1997; Kolodner 1993): Knowledge is stored in the form of cases-solutions. 3. Rule-based format (Triantaphyllou & Felici, 2006): If-Then rules are stored as the knowledge source. 4. Frame-based format (Minsky, 1975): Objects are divided into several frames, and each frame contains its corresponding attribute to describe the characteristics of the objects. 5. Ontology (Munn, 2009 ; Uschold & Gruninger, 1996): It is a representation of some pre-existing domain of reality which reflects the properties of the objects within its domain in such a way that there obtains a systematic correlation between reality and the representation itself. It is formalized in a way that allows it to support automatic information processing. 22