Research Of Geospatial Semantic Integration And Semantic Similarity Measurement Based On Knowledge

Our society has been in the ambient information age dominated by Big Data.The more convenient of obtaining data and greatly enriched data significantly are changing the structure of human society and our understanding of the world.In the era of big data,the range and depth of requirement of interaction,between human,machines or human and machine,are increasing and deepening.And the demand for integrated application of different data is becoming more and more urgent.However,while the capacity for data acquisition is growing rapidly in this era,we seem to be trouble in an embarrassing situation that "the data is large but small".The main reason is the weak of the knowledge sharing ability which is caused by the heterogeneity of multi-source data.This hinders the efficient and cooperative utilization of data.The heterogeneity,inconsistency,scale etc.of data have been the great challenges of the analysis of big data.The essential cause of these challenges is the semantic heterogeneity bringing about the problem which is more prominent and the contradiction that is more urgent in GIScience.Therefore,this thesis focuses on the study of geospatial semantic.This study mainly analyzes the semantic of geospatial semantic,and proposes an implementation approach to the integration of geospatial semantic and an algorithm of measuring the geospatial semantic similarity.This thesis includes five main contents as following:(1)The thesis summarizes the research areas of geospatial semantic,the progress of integration of geospatial semantic,and the study of geospatial semantic similarity measurement.It also points out that the semantic of the term "geospatial semantic" is ambiguous in many researches,which means "geospatial semantic" sometimes represents completely different meanings in various circumstances.And the research of geospatial semantic integration which declared to dedicate to the semantic study actually is the research of grammatical structure.Based on the problems aforementioned,the main contents and the structure of this study are briefly described.(2)This study has analyzed two aspects of the meaning of geospatial semantic,which are the understanding and the formalization of geospatial semantic.And the understandings of geospatial semantic based on knowledge and data are distinguished.Based on these two different ways of semantic understanding,differences and relationships between map semantic and image semantic are emphasized to discuss.Following this,three different forms of representing semantic are discussed based on knowledge.Our main focus is to introduce ontologies that represent the semantic.The study also introduces the formal language OWL which describes ontologies and a popular tool that is used to build ontologies.And a simple inland hydrological ontology is built using this tool.(3)The study has proposed a new approach of integration of geospatial semantic.The mathematical tools including order theory,lattice theory,formal concept analysis,rough set theory and information entropy etc.are introduced in details.According to the needs of this study,some definitions are redefined.An approach of building and reducing a combined weighted concept lattice is proposed based on the inclusion degree and information entropy,and the reduction of the lattice is analyzed from the perspective of the granular computation.Following this,the experiment of integrating semantic is implemented whose experimental data are extracted from GB/T 13923-2006(national specifications for feature classification and codes of fundamental geographic information and SL 213-98(specification of basic information coding of water conservancy projects).And the reduction results of combined weighted concept lattice under various thresholds are analyzed.(4)A new approach which is used to measure the semantic similarity is proposed.The research status of semantic measurement is summarized.The review focuses on discussing various models and approaches of semantic similarity between concepts based on knowledge.According to the structure of lattice,the relative hierarchical depth of the lattice is defined.Combining the relative hierarchical depth and the feature-based model and information theory model,a hybrid approach is designed to measure the semantic similarity of a pair of concepts.Using the experimental data extracted from GB/T 13923-2006(specifications for feature classification and codes of fundamental geographic information)and GB/T 20258.1-2007(data dictionary for fundamental geographic information features),an experiment is implemented to measure the semantic similarity of concepts.At last,by comparing the experimental results from the feature-based approach,the validity of this hybrid approach is evaluated.(5)An experimental platform is designed to implement the integration of geospatial semantic and semantic similarity measurement.The functions of this platform include data transformation,computation of the combined weights of the attributes,generartion of weighted concept lattice,plotting of concept lattice and calculation of semantic similarity.The experimental results of building weighted concept lattice,reducing weighted concept lattice and calculation of semantic similarity are demonstrated.