Research On The Ontology-oriented Automatic Semantic Classification Of Geographic Information

With the continuous development of social informationization, the traditional concept for GIS as Geographic Information System has been expanded into Geographic Information Science and Geographic Information Service. The booming development of theories, technologies and applications has caused the globalization of GIS, which means the subject is executable, open, distributed and networked. The urgent need for sharing and operating geographic information has been induced by globalization of GIS, the research of Geo-ontology has been developed rapidly, and a realistic significance of the semantic of geo-spatial information research has been shown to us. While many great breakthroughs took place in this field, there are still several to-be-solved problems:(1) the rationality and reality of geographic information classification; (2) the analysis of the complicated semantic relations of geographic concepts; (3) the degree of Geo-ontology's automation and openness is relatively low.Aiming at the problems above, this dissertation uses the research of geographic information semantic classification based on ontology as a starting point and conduct formula concept analysis by extracting the attributes of geographic ontology automatically, proposing a method that achieves the automatic classification of geographic semantic information. Based on theories of the geographic ontology, the formula concept analysis and the natural language processing, this dissertation focusing on studying the source of ontological attributes of geographic concepts and some methods which include formula concept analysis in the classification of geographic semantic information, Chinese word segmentation methods in geographical information science, automatic annotation and extract method of the ontological attributes.The main research work is divided into parts as follows:The research status of geographic semantic information classification is represented and the main problems, which will be solved by the research, such as the rationality and reality of geographic information classification, the richness of semantic relations of geographic concepts, the openness and automation of constructing geographic ontology, are pointed out. For these problems, this dissertation put forward a strategy to achieve the automatic classification of geographic semantic information by automatic extraction and formula concept analysis of the ontological attributes of geographic concepts.With the computer technology developing, many deficiencies arise in the geographical information classification scheme. This dissertation compares various existing geographical information classification scheme, and it tries to improve the classification schemes by geographic semantic classification based on ontology and forms a geographic semantic information classification scheme through formula concept analysis. By studying the habitation, land and rivers in GB/T 13923-2006, it proves the feasibility of this strategy both in theory and in practice. It also analyze the classification results and compares them with that worked out in the traditional way.For the problem that some ontological attributes needed to be manually determined in the past, it proposes a strategy that extracting ontological attributes and annotate Chinese words automatically by defining description sentences for geographic concepts. And it studies the phrase segmentation method which concerns to GIS, providing effective data infrastructure for the ontological attributes annotations. It also suggests analyze the results by much more semantic relation type, which will enrich semantic relations and help the classification of geographic semantic information.The dissertation analyzes the fundamental method to extract ontological attributes from description sentences, and reveals a mapping relation between words in description sentences and ontological attributes. An entity recognition mechanism of information extraction is applied to word recognition in description sentences, thus achieving a conversion from unstructured natural language sentences to structured statements which are annotated as ontological attributes. With GATE semantic tagging plug-in, a formula concept table is produced by constructing word table and JAPE rule of ontological attributes, and the results is analyzed and compared with the artificial results.On the basis of the theories and methods being researched, the dissertation concludes an overall framework of the geographic semantic automatic information classification system, which includes functions such as Chinese word segmentation, ontological attributes annotation, formula concept semantic classification and so on. It obtains a semantic classification hierarchy through tests over geographic concepts, and realize a tentative attempt in the automatic classification of geographic semantic information.The research on the ontology-oriented automatic semantic classification of geographic information aims at enriching the theoretical system of constructing geographic ontology on a classification level. With the theories and methods concerning to geographic ontology and natural language processing, the research has something enlightening for the rationality and reality of geographic information classification, the analysis of the complicated semantic relations of geographic concepts and the automatic degree of semantic classification.