Researches On Semantic Annotation Technology For Marine Literature Metadata

Nowadays, marine literature metadata is developing fast with the flourish of marine science. Different marine subjects are inter-crossed in the research area; marine literature metadata of different subjects are heterogeneous. Therefore, solutions for the share and interoperability of marine literature metadata are urgently needed. Semantic metadata can describe semantic information of resource, and support data share and interoperability in the semantic level. Semantic annotation technology can translate metadata into semantic metadata. Based on this ground, the thesis investigates semantic annotation technology for marine literature metadata.From our research and analysis of formerly proposed semantic annotation tools and methods, this thesis shows that semantic annotation technology for marine literature metadata is characterized by two key technologies: the technology of Automatic Marine Literature Metadata Selection (AMLMS) and the technology of Automatic Semantic Annotation for Marine Literature Metadata (ASAMLM). Marine literature metadata can be classified and selected automatically by using AMLMS. Meanwhile, the ASAMLM then can automatically translate the chosen metadata into semantic metadata. The two technologies and their implementation are the focus of the thesis. The AMLMS is based on the theory of machine-learning-based text categorization.This thesis compares the most three famous classifiers including maximum entropy model (MEM), support vector machine (SVM) and Adaboost. The result of comparison experiment shows that the best classifier for AMLMS is the maximum entropy model whose precision and recall reach 99.2492% and 94.4286% respectively. The automatic classification system for literature metadata is implemented with C#.This thesis deeply analyzes XML Schema, and finds that there are much semantics of the domain knowledge inhering in the structure of metadata. Base on this investigation, the thesis presents an algorithm for automatic generation of ontology, which extracts the semantics and generates original domain ontology automatically by parsing XML Schema. Further more, this algorithm can generate semantic mapping between the structure of XML and ontology. So this thesis proposes a new method of automatic semantic annotation for metadata. The method gets the semantic mapping by the algorithm, and annotates metadata automatically with the guidance of the semantic mapping. This method can be widely used for the metadata defined by XML Schema, and has wider usage than GRDDL. The method is implemented with Java and Jena.The technologies of semantic annotation for marine literature metadata are universal and applicable for metadata in other field.