Research Of Ontology Matching Key Techniques

Into the 21th century, the embryonic form of the knowledge society has appeared. Not only used by the enterprise in the field of production, from a macroeconomic point of view, knowledge also play a crucial role to the prosperity of the community. As knowledge becomes the core of economic resources, knowledge representation and management has become imminent. Moreover, in front of a large collection of information, people pay more attention on finding a way to integrate the knowledge of these different methods together. Using the ontology which detailed and clear, and contains a wealth of semantic information, can express this purpose. Creating an ontology is to achieve cross-cutting, cross-task knowledge sharing, reuse and reasoning. The same time build a model for static information used ontology, the general reasoning mechanism will be set up the necessary physical components to be integrated into reusable components for the new knowledge system.Different applications for their own purposes, different areas define a number of ontology. But some ontology have expressed similar content. These ontology have considerable differences in name, or the structure, this is the ontological heterogeneity. This problem exists in a variety of applications related ontology. In order to build synergy interoperability between the agent or service using different ontologies, ontology mapping is the best way to solve this problem. Ontology mapping can identify the semantic relationships between different elements of multiple ontologies. Therefore, it can solve the multi-agency or service coordination issues. And the user can use a transparent manner to achieve the knowledge access in a number of ontology, and then complete the task of searching or browsing the applications. And merging the information of a variety of ontology will provide greater harvest. Therefore, ontology mapping became a key aspects in building an integrated knowledge network. Currently, the ontology mapping is not only an important research topic in ontology engineering, information integration, Web service composition and peer-peer applications pattern, but also a very active research area.This article focuses on ontology mapping problem, for different aspects of the problem, launched a series of studies, and proposed four different types of ontology mapping algorithm to solve a variety of practical problems. First, for the demand of large-scale ontology mapping, this paper propose an ontological compression algorithm based on the concept of cluster. Concept is a critical component of the ontology, the other components are directly associated with concepts. Therefore, when compress the ontologies in the mapping task, our approach based on the concept, according to the semantic relationship between concepts, use concept of clustering and compression process, namely:remove the concepts that have nothing to do with mapping; stop until the number of clusters obtained consistent with the threshold value. Objective is to get a reasonable reduction to the volume of ontologies involved in mapping.According to the special role of concept in the ontology, this method use DICE coefficient method to measure their semantic similarity. Subsequently, by calculating the entropy of properties, the method gets the semantic similarity of concept. And integrate the concept of semantic similarity and semantic similarity to get semantic relationships between concepts. Finally, according to the semantic relations of concept, and through the concepts clustering process in space concept, this method completes to compress the size of ontologies. That is by the iterative way, remove the concepts that have nothing to do with mapping gradually, and integrate the concept of a higher semantic similarity as a group. Discover out the concepts which are meaningful to mapping and compress them into concept clusters, then model and solved that problem.Second, for the n:m mapping problem between ontologies, we propose a ontology block matching based on clustering method. According to the basic idea of the clustering:classify the data samples by the distance between them, so that the data in the same class with the greatest similarity, the similarity between different categories is low. block on the ontology, do the condensation process continuously between the class clusters. When the blocking process finish, each m:n mapping between the entities within the cluster can be found.Firstly, computing semantic similarity between matching element by the resources in Hownet. Then, block the ontology entities based on the idea of hierarchical clustering; first of all put all the entities of the ontology in a space, each entity as one class cluster, then constantly look for the nearest two cluster classes by the semantic distance, merge them into one class cluster; next, the number of clusters reduced by one; until the remaining number of clusters and pre-class set of k values are equal, block to stop. This approach allows complete block and mapping the same time, reducing the complexity of the block mapping problem. In addition, the use of computing semantic similarity by the resources in Hownet can also solve the problem of Chinese ontology mapping.Again, this article proposes a hybrid ontology mapping method, considered the various features of ontology. The mapping of ontology element level and structure level can join together. Methods use the tool based on linguistics, distance to calculate the element name similarity, and use this similarity iteration method on the semantic similarity graph calculate the structural similarity, and through constant iteration to adjust the value of the similarity results. Thus, not only to achieve the measure of the matching relationship between name or structure of the ontology, and allow an appropriate way to mix the two forms of mapping. This method attempts to a new map way, namely:use the tool of semantic distance and character distance point out the mapping relationship of ontology element level, and build the semantic graph, the element level and structure-level mapping that combine to complete the problem solved. This avoids using a single method can not be less than all of the information ontology, problem solving for the map provides a framework to map the results more desirable.Finally, the similarity between ontology entities, mergers and extraction, or the weighting method used, or direct combination method requires the participation of a large number of artificial problem, a support vector machine based ontology mapping method. This method proposed a new similarity combined the idea of the concept put forward a similar cube, through its on the cut, the problem will be mapped into that classification. The help of support vector machine to complete this task. Thereby reducing manual intervention and improve mapping accuracy.Methods used in combination with Wordnet and Jaccard Simlarity the concept label similarity method; put forward the concept of neighbor levels, through the neighboring layers to reveal the body element of the concept of structural similarity; use vector space model method and examples found in Ontology similarity between attributes; Presented the results of a new multi-strategy combined similarity method, proposed a similar concept of a cube, cut through the operation, taking a similar vector of support vector machine-based mapping discovery strategy, that task will be mapped into a vector space II Value of the classification. Problem solving in reducing the degree of human intervention and improve the quality of the mapping.Although the common data set on the experimental results show the effectiveness of these methods, but they also exist some defects and problems, such as:Some of the performance of mapping methods rely on pre-set parameters and threshold value adjustment; some Method can handle a single type of data comparison; some methods require the input of the body content can not be too rich. Therefore, in the next step of work, these issues will be targeted improvements to further enhance the performance of the method.