Research On Semantic Annotation For Periodical Literature In Multi-granularity Way

The process of tagging target resources’concept classes, properties, instances, and other metadata are called semantic annotation. A typical semantic annotation process is to input documents and ontology in annotation system, then semantic tagging system get information from the body, and to add semantic information in the target document, finally output the marked document. After semantic annotation, the document can be upgraded from machine-readable condition to the status that the semantic information can be understood by a computer.Based on the degree of artificial participation in the process of semantic annotation, semantic annotation methods can be divided into artificial semantic annotation, semi-automatic semantic annotation and automatic semantic annotation. Manual and semi-automatic annotation are accurate, but in the case of today’s explosion of information resources, they have shown significant drawbacks that they are time consuming and inefficient, and cannot be fully realized by artificial mark in large quantities of semantic annotation work. Therefore, the research on automatic semantic annotation has more and more attention. Semantic annotation is the basis of semantic web technology, and automated semantic annotation has become a major bottleneck in the development of the semantic web.This thesis describes the background knowledge of the semantic web, including the basic theory, architecture, the differences between semantic web and the World Wide Web, ontology theory, the research status of the semantic web and semantic annotation, the relation between semantic annotation and ontology, and analyzes the structure of journal articles, and design a semantic annotation method for its characteristics, and the journal articles are divided into three level of coarse, medium and fine granularity, semantic annotation were made for different levels of granularity.Aiming at the problems of automatic semantic annotation, the proposed multi-granularity semantic annotation for papers include the following steps:(1) obtain data document of each granularity of papers; (2) extract user dictionary from the ontology, add it to Chinese word segmentation system; (3) identify the concepts, attributes and examples in different size document based on the establishment of a custom dictionary; (4) assemble identified concepts, instances, attributes and attribute values using RDF framework, and form RDF triples; (5) form RDF documents on the basis of assembling RDF triples.Finally, in order to verify the effectiveness of semantic annotation method proposed in this paper, multi-granularity semantic annotation experiments is designed for journal literature. In the experiment, for the journal literature in a certain field, multi-granularity data are acquired, concepts, attributes and examples are identified, finally RDF triples are assembled, and RDF document are formed. The results show that the proposed method can be used in semantic annotation for journal articles in a certain field, and it is more efficient than artificial semantic annotation, but also can ensure labeling accuracy in a certain degree.