Research On Several Key Technology Of Ontology Engineering

Ontology engineering is a set of activities on ontology, including ontology organisation, maintenance, management, domain knowledge ontology design, building, modeling methods, knowledge sharing and reuse, and evaluation processes etc. Ontology construction, ontology mapping and ontology evolution are three key technologies for ontology engineering research area; the main work of this thesis is to focus on this three points.Ontology construction is the basis of the ontology engineering, although many of the currently existing ontology, however most of them are designed for specific areas, which is built with the support of domain experts and the way through the manual set up. The principle for building standard ontology includes the following five issues: (1) clarity and objectivity: The ontology should be natural language, a clear description of the effect of the content of defined terms, definitions must be objective, and the background is independent, define a complete and clear as possible, all the definitions should be described in a natural language; (2) Complete: the definition given should be complete and fully described to express the meaning of the term; (3) Consistent: the terms inference and the meaning of the term itself is compatible, not a contradiction; (4) Maximum extension: add to the ontology in general or specific terms, you can not modify any existing content, that ontology should be able to post some of the tasks can be expected to provide the conceptual basis; (5) Minimum commitment: dealing with the modeling object with the constrains as little as possible .In the deep study of the current popular method of ontology construction, the author proposed a method for extract ontology from the relational database. In this part of the work, analyze the advantages for using relational database as the data source, the advantages including the follow issues(1) areas of relatively high correlation (2) The relationship between the relational database tables using the form of two-dimensional data storage, databases, message format within the neat, easy to automate the analysis and extraction. (3) Relational database model in fact implies a conceptual model for the related fields. Based on this analysis, given a collect of rules for extract ontology from the relational database, including how to get the concept from relational database;How to obtain the concept properties and set property's domain and range; how to assign the relationships between concepts; how to get the instance for concepts; how to classify the concept base on the table contents; how to apply the integrity constrains of relational database as the constrain within the ontology. Due to the deficiencies of synonymous concept and relational hierarchy within the relational database, and WordNet in these two areas have special advantages, where the generated ontology using WordNet made further improvement. The use of the existing data source ontology extraction method of automatically reducing the intensity of manual operation and improve the efficiency of ontology construction.Ontology mapping is put forward is in order to solve the heterogeneous problem of different ontology. The reason for ontology heterogeneous, mainly because most of the existing ontology are based on needs of different projects, and the the ontology is build with different guiding principle. Structure heterogeneous is mainly seen in the following two aspects: (1) concepts or attributes of different granularity, leading to structural isomers. (2) by the same concept, if not the same as used lead to different dimensions of the concept of tiered system, will lead to the heterogeneous structure layer. In order to achieve interoperability between heterogeneous ontologies, must address this problem among heterogeneous, generally there are three ways for solving this problem, the first method is: to establish an inclusion relations between ontologies; Method Two: establish a mapping between ontologies; Method three: build a public ontology. Among these three methods, the mapping between ontologies is the most effective one; the purpose of mapping is to find out the relationship between the concepts of ontology, based on these relationships to develop appropriate mapping rules. An ontology mapping process should include five sections, which are: (1) standardization of ontology; (2) the similarity calcultation; (3) semantic mapping (4) the implementation of mapping; (5) mapping post-processing; the extraction of the similarity which played a decisive role, it's main aim is the similarity calculation, and then to determine the similarity mapping. Semantic similarity between concepts and the semantic distance between concepts is closely related. In general, the semantic distance is a range between 0 and infinity a real number. If the semantic distance between concepts is greater then their semantic similarity will get lower; a concept with its own semantic distance is 0; when the semantic distance between tow concepts is 0, then their similarity is 1; if the distance between two words is infinite, their similarity is 0; if semantic distance between two words going greater then the similarity of them will be smaller. With the study of some current hot ontology mapping method, found that most of these methods map all the concepts in the ontology together, due to there are many irrelevant concepts involve the mapping process, because these are not related calculation of the concept map mapping algorithm greatly reduces the performance and accuracy. This part of the study, the authors propose mapping method, the method based on the concept of hierarchical classification of concepts, and top-down analysis of concepts. This method is greatly avoiding the calculation similarity of irrelated concepts, and reduced the computational work. This mapping algorithm improved the mapping performance and get accuracy result.Ontology evolution is a process of self adaptation for changing. The change of ontology always caused by the following three reasons; including the changes in the domain knowledge, the conceptual model change and represention changes. Ontology evolution ontology is a process for manage the modified ontology, after the modification taken place, the ontology which associated with this should take appropriate action, that make the ontology can maintain semantic integrity, data accuracy, to ensure completion of application requirements.The most import influence may lead by ontology evolution is inconsist and incompatible problem. It means that the ontology after change can not meet the original requirement, and can not fit the designed function. So the most important issue for ontology evolution is deal with inconsist and incompatible problem. A complete evolution process includes six stages, including the change capture, change representation, semantic change, change spread, change implemention and change confirm. These six stages are not independed from each other, they are intersected. The first two steps are providing the requirement for guiding the evolution; others are core part for the whole evolution process. For the research on ontology evolution, this thesis surved the existing evolution algorithms, it can be seen, that most of evolution method still done by manual work, it is very time consuming and complicated. There some ontology edit tools can support the evolution, however, they always use deep delete method, it will remove too much useful information within the ontology and make the ontology can not meet the requirement. This thesis proposed a evolution algorithm which base on the Cost calculation constrain and applied the heuristic graph search method to control the evolution process. Here, the author gives the definition of evolution cost first, by analyzing the entities contribution for the ontology, represent the method for calcultation the Cost for different operatin apply on the ontology entity. Put forward the definition of minmam qranularity associate operation and prove they can complete all evolution needs. The evolution operation strage is designed to simplify the evolution process. Finally, apply the cost calculation and operation strage to heuristic graph seach, this method add the constrain for graph seach and it reduce the time consume and memory consume.Currently works on ontology is still attracting research experts and scholars. This thesis'work is expected to play in this field to promote the development of the role. In this paper, the proposed method has a high theoretical and practical value. Hopfully, it will be benefit the future research work on ontology engineering research area.