Research On Improved Attribute Exploration Algorithms And Ontology Construction

With the rapid development and applications of Semantic Web, we need to search information from the mass data. An ontology, which is also the basis of Semantic Web applications, is a formal specification of a shared conceptualization. Therefore, the coverage ratio and precision ratio of the web search are directly impacted by the quality of the ontology construction. Nowadays the main ways to construct ontology are manual or semi automatic. Because of the differences on the constructors and living environments, it happens that different people will get various versions, even when they are constructing the same ontology. As a result, the finding of better methods on ontology construction is always the hotspot and a difficult problem for the experts and researchers on the Semantic Web.In the face of the massive data information, we need a valid way to extract, organize, and represent the information, and provide the reference models and hierarchical relations for constructing ontology through computing. The formal contexts and concept lattices in formal concept analysis are good methods of knowledge representation. The attribute exploration algorithm, based on the formal context, can acquire the maximal knowledge hidden in the formal context, by asking a domain experts successive questions, finding implication between attributes which can express the knowledge on inclusion between object sets, and get all of the intent and stem base, and provide a valid way to resolve the problems of ontology differences and ontology completions on the ontology construction. The attribute exploration algorithm not only can be uesd in the determined formal context, but also finds the intent, pseudo-intent, Duquenne- Guigues base in the uncertained formal context, infers the implicit knowledge in the domain, finds the relatioships between each attributes, and the corresponding relationships between attributes and objects. It resolves significantly the problems when constructing ontology in the uncertain situation, and improves flexibly to expand the formal context to be completed. Therefore, the research on the attribute exploration algorithm and its improvements, whose conclusions are applied to construct ontology in order to propose the new ontology construction methods based on attribute exploration algorithm, is of great significance on theoretical and application value in the future.The innovation of this paper is that the our theoretical research on the attribute exploration algorithm is pioneering in domestic, and we have shown that the existing algorithm in the literatures has many redundant computations in different situations, then propose two improved algorithms, comparing these three algorithms, finally using this conclusion to construct the concept lattices and combine ontology construction, proposes new ontology construction methods.The contents of this paper are organized as following:Firstly, it detailedly introduces the basis theoretical knowledge on description logic, formal concept analysis, ontology and so on. The research on the intent, pseudo-intent, Duquenne- Guigues base is pioneering. After lucubrating the settings on every technical detail and all the computing process, we find the redundancy in the course of computing the next Bi+1, and accounted for its possibilities, providing the theoretical foundation for ontology construction.Secondly, it proposes two improved attribute exploration algorithms. According to the properties and characteristics of implications, the improved attribute exploration algorithm (Ⅰ), by means of finding the next attribute set Bi+1 in the lectic order and deciding it is whether intent or peseudo-intent in terms of relevance, avoids the case of redundant computing, and proves its completeness. In the case of the cardinality of attribute sets on the small side, the improved attribute exploration algorithm(Ⅱ) lists all of the subsets of attribute sets, then splits the sets in batches according to the same cardinality of attribute sets, and inspects their relevance group by group. Through comparing and analysing the differences among the primary attribute exploration algorithm and its improved algorithm(Ⅰ) and (Ⅱ), it accounts for their respective advantages and disadvantages and corresponding applications. The improved algorithms not only simplify the judgement conditions, but also improve the efficiency of the algorithm.Thirdly, considering the attribute exploration algorithm as the link, which contacts the ontology construction and formal context, we proposes an ontology construction method based on attribute exploration algorithm of description logics(AEOCM): via extracting the formal context from the data sources and attribute exploration, generating the concept lattice, transforming the lattice to the ontology according to the transformation rules, which is acquired by comparing and analysing the differences between ontology and concept lattice, after adding and modifying the ontology manually, finally finishing the work of ontology construction. So we could make use of the above method to construct and complete description logic knowledge base.Fourthly, using the method of AEOCM, we have constructed the ontology about the two squares location relationships. The computing situations of every section are introducted in detail, in manner of instantiation. AEOCM has risen from theoretical level up to practical application level. In this paper, two improved attribute exploration algorithms are proposed, after going into and analysing the primary attribute exploration algorithm; and these algorithms are applied to construct ontology. As a result, we propose an ontology construction method based on attribute exploration algorithm of description logics (AEOCM), and specify the implementation procedure of this method in terms of instantiation. It provides a feasible, workable and effective way to construct ontology.