Font Size: a A A

Linked Data And Knowledge Representation Automatic Semantic Annotation

Posted on:2013-11-13Degree:DoctorType:Dissertation
Country:ChinaCandidate:M XieFull Text:PDF
GTID:1228330395975991Subject:Computer system architecture
Abstract/Summary:PDF Full Text Request
The unstructured, semi-structured quality curriculum data is on a growing trend. The traditional manual or semi-automatic semantic relationship extraction, labeling gradually become the bottleneck of data construction engineering related courses, automatic semantic annotation has urgent practical needs. In this paper, in-depth study of the associated data at home and abroad, knowledge representation, semantic annotation results of automated semantic annotation strategy oriented semi-supervised mass network resources. Explore the conceptual content and the relationship between heterogeneous networks in knowledge representation, semantic mapping, loaded mechanism of complex and diverse relationships and deep body automatically generated mechanism.In this paper, the automatic semantic annotation planned four-layer model, bottom-up, respectively, for data extraction and conversion layer, data correlation and processing layer, data storage and indexing layer, ontology construction and annotation layer.Data conversion layer is responsible for the document is converted to RDF triple. Curriculum document disambiguation based on existing or new body or associated data queries LOD, if there is a corresponding entry, class mapping entity associated attributes associated output RDF triple. If there is no corresponding entry in the semi-supervised way to automatically extract concepts, properties, instances, and the concept of the relationship between, said the course subject, object, predicate form data into the machine and the person can understand RDF data. Data processing and the associated layer is responsible for processing the RDF triple associated data. Query LOD associated open data sets assigned a class label from the correct body after normalized RDF data to be mapped with the corresponding entity, outside in the form of the chain to generate a new association data. Associated data analysis, and found that the potential relationship between entities, will find the relationship between semantic automatic annotation, related courses data generated after processing.Data storage and indexing layer is responsible for the storage and indexing associated data. According to the size and speed of the associated data requirements, local RDF file different forms of the MySQL database, cloud storage, storing data associated with the main body of the triple RDF associated dataset, predicate, object indexing. Ontology construction and annotation layer is responsible for the body built on top of the associated data, to store and index related course data through semantic annotation technologies such as, increase TBOX associated data further to become the body. Ontology construction is completed, according to the needs of different types of documents, automatic ontology tagging algorithm, automatic ontology annotation. If the file type is OWL or RDF, directly corresponding to the related items in the existing body, automatic ontology annotation. If the file type is a resource document were to invoke the concept of recognition, class identification algorithms relationship identification operation to complete the automatic ontology annotation.This paper designs automatic semantic annotation experimental system algorithm. System function and use through the application examples. Data quality courses as experimental subjects, and analysis of automatic semantic annotation entity mapping accuracy, the relationship recognition accuracy rate of automatic semantic annotation experiment, three groups before and after the sequence contains, SameAs relationship automatic annotation accuracy. Compare experimental system for automatic semantic annotation with existing automated semantic annotation system Text2Onto, Pretege, Gate Chinese entity mapping. Massive network data to meet the needs of automatic semantic annotation. According to the experimental results, the automatic semantic annotation entity mapping accuracy rate of63.27%, the relationship recognition accuracy rate of up to37%. In automatic semantic annotation experiment data of the associated courses the three groups relations before and after the sequence contains, SameAs automatic annotation accuracy respectively82.63%,80.57%,92.95%. Chinese entity mapping experimental system accuracy rate of92.93%, recall rate was82.55%. Higher than the existing Text2Onto, Pretege, Gate automatic semantic annotation system. Automatic semantic annotation of a massive network data can be used as a new way.
Keywords/Search Tags:linked data, knowledge representation, automatic semantic annotation
PDF Full Text Request
Related items