关键词: 领域本体, 自动构建, 知识管理, 信息管理, 文档概念图, 频繁子图挖掘, 信息子图, 算法

Abstract: Aiming at the shortage of domain ontology component algorithm, a graph-based approach for automatic construction of domain ontology oriented to knowledge and information management was proposed in which concept extraction and relationship extraction were included. Each document in the collection was mapped as a document graph. Random walk term weighting was employed to estimate the importance of the term information to the corpus from both local and global perspectives. Graph vertex clustering algorithm was used to classify terms with different meanings and group similar terms to generate candidate concepts. An improved frequent sub-graph mining algorithm constrained by both vertices and information was proposed to find arbitrary latent relationships among these concepts. For ontology evaluation purpose, a method for adaptive adjustment of concepts and relationship with respect to its practical effectiveness was brought forward. Finally the domain ontology was formed by describing domain concepts and relationships using OWL-DL. Evaluation experiments showed the effectiveness of this algorithm.

Key words: domain ontology, automatic construction, knowledge management, document graph, frequent sub-graph mining, informative sub-graph, algorithm