Acquiring Attributes Knowledge Based On Semantic Grammar

Knowledge acquisition is considered to be the bottleneck of knowledge engineering and artificial intelligence. Acquisition of attribute and attribute-Values of concept is one of the most fundamental research direction in the field of text knowledge acquisition. Concept is a special term which has certain semantics. However, it is difficult for the concept to directly express the semantics. So we need other type of knowledge to explain and describe the concept. And the semantics of the concept became clearer after further express.Attributes of concepts are exactly one kind of such knowledge.Attributes describe the most essential characteristics and nature of a concept, which are able to identify or describe the concept. Therefore, the attribute and attribute-value acquisition plays an important role in the field of text knowledge acquisition.Previous research on the attribute knowledge acquisition had been done much. But it ignored the condition for establishment of attribute-value, implicit attribute and relational attribute, and there is no structured representation about acquired attribute knowledge.Additionally, the number of acquired attributes is limited, and weigh too heavily toward verifying common attributes. It ignored the immensity and diversity of a real attribute space.To avoid the shortcomings, a method to obtain attribute knowledge is proposed, which depends on attribute semantic grammar. The main works of this paper include:(1) Constructed attribute semantic grammar which can match attribute knowledgeThe design of attribute semantic grammar is vital for researching the text of attribute knowledge. First, matched natural language texts which contain attribute knowledge after marking up large corpus, then analyzed the law and mode of the attribute knowledge and constructed corresponding attribute semantic grammar. All the semantic grammar form a pattern base, each semantic grammar can match natural language texts which contain attribute knowledge.(2) Extracted the attribute knowledge in the form of the predicateMatched text corpor with the help of the EKEL parser and attribute semantic grammar,and get the parse-tree. Then, extracted the attribute knowledge in the form of the predicate from the parse-tree, and it structured the text knowledge. This method can also be used to handle a lot of special attributes. The form of the predicate can also solve the problems of structured expression of the condition for establishment of attribute-value, implicit attributeand relational attribute.Experiments indicate that with the attribute knowledge acquisition from text in the Internet, the method has achieved a precision rate of 70.86%, and better than the traditional method of lexico-syntactic.