Automatically Semantic Annotation of Network Document Based on Domain Knowledge Graph

Abstract

Massive network document resources provide abundant retrieving and reading information, but it is consuming and exhausting to quickly search, understand and analyze those documents. In order to seek semantic support for searching, understanding, analyzing, and mining, this paper proposes a more convenient way which based on domain knowledge graph to annotate network document automatically. The method firstly adopts an upgraded TF-IDF model based on the contribution to quantify instances in knowledge graph, then analyzes the semantic similarity between unannotated documents and instances based on Jaccard distance and lexicographic tree distance comprehensively. After the accuracy tests conducted by collecting network documents, the results show the initial marking accuracy is up to 74%, successfully certifying the method being able to automatically annotate network documents in terms of semantics from the domain knowledge graph.