The Construction Of Aviation Knowledge Base Under The Entity-attribute Frame

As a basic resource, knowledge base is essential for all kinds of information processing systems. However, the relatively well-known knowledge bases are mostly constructed by artificial means or need the participation of domain experts, and automatic knowledge base construction is still a difficult problem.To solve this problem, this paper proposes a method of automatically constructing an aviation knowledge base based on the entity-attribute frame, in which knowledge is expressed in the way of natural and structuralized frame after extracting each attribute of the entity, and applied to information retrieval system, and in which the original query keywords are expanded effectively. As a result, it improves the retrieval performance. This paper contains the following sections:Firstly, the paper focuses on knowledge acquisition which is the key technology in the process of knowledge base construction. According to different types of knowledge, such as the relationship between entities and attributes information of the entity, this paper introduces an SVM-based entity-relationship classification method and a multi-strategy based attribute acquisition method. The former uses the interpretation items of two entities as pseudo-documents, uses an SVM classifier to make a class decision, and thereby classifying the entity pairs into one of eight entity relationships which are artificially defined in advance. The latter uses the resources that cannot be fully utilized by the above-mentioned SVM method, and extracts the attribute information from the interpretation of the input items by using rules, templates, etc. Finally, the acquired knowledge is added to the entity-attribute frame to construct an aviation knowledge base in a certain scale.Finally, because the traditional keyword-based retrieval technology has a shortcoming of simple morphology matching, this paper proposes a method of using the constructed aviation knowledge base to expand the original retrieved keywords to a set of semantically similar keywords so as to form a secondary retrieval. Thus, the retrieval performance is improved, and better retrieval results can be obtained.