Research On Ontology Mapping Methods

As the Internet information resources are increasing rapidly, the network application anddemand is also continually expanding. Traditional Internet technology does not consider theorganization structure among network resources in spite of completing the connection ofnetwork resources, with the various knowledge in disorder and thousands of storage media inscattered distribution. How to find user-needed information accurately and quickly in the vastinternet resources become urgent issues to tackle. In order to make the interoperabilityavailable between different Web applications and services, ontology semantic heterogeneity isthe crux to solve the problem. What falls into our domain of interest is an ontology existingcomprehensibly as the machine language and the simplification and abstraction expressed.And the most important application domain of ontology is the semantic Web and the problemof the semantic description and ambiguity of semantic Web can be well solved by theontology.But distributed patterns and heterogeneity are characteristics of semantic Web, andontology description of the domain knowledge can’t be shared because there also existisomerism and inconsistency even if it is the same field of ontology. Therefore, solving theheterogeneity between ontology becomes a key priority and ontology mapping becomes thehot research topic.However, although there are many relevant mapping and matching method, but there hasnot been a clear and successful scheme which can fully adapt to the future development need,and can not complete all the operations automatically without any domain experts. Therefore,many unsolved questions still exist. Based on the problems existing in ontology mappingmethod in a series of research, this article puts forward several related methods in order tosolve the practical problems.First of all, many mapping algorithm were too dependent on ontology of the entity in thestring information and body structure. These techniques can sometimes get very goodmapping results, but sometimes mapping fails. Using the example to enrich each conceptnode’s semantic information is the most effective way to deal with this problem. Mappingmethod based on the example uses the example’s corresponding "text" appearing in the wordand its frequency to find out the mapping relationship between elements. The richness in theexample information of this method determines the mapping efficiency. Ontology of the entitymay have multiple examples, each instance contains instance name and its associated attributevalue. But many ontologies under construction do not add some example information for eachentity, which will directly influence the performance of the strategy based on examplesmapping. In addition, uncertainty problems in mapping remain to be solved.This paper proposes an ontology mapping method based on extended information. First,use of information retrieval based on ontology methods will expand ontology text as anextension information about ontology concept and attribute node. Then, the ontology as classification tree, the use of source ontology examples of information as a training set, theuse of text classification method based on hierarchy structure in ontology concept andattribute node classifier, and goal ontology examples can be used as a test set to sourceontology of node classification. The purpose of the work is to get the equivalence betweenentities and inclusion relation of probability model. Finally use the ontology mapping methodbased on probability theory for mapping set. This method can extend example set forno-example information ontology; improve the effectiveness of the mapping method based onexamples, by combining theory of probability mapping method. That can get not only theequivalence relation mapping, but also the inclusion relation mapping, to a certain extent,solve the uncertainty problem of found mapping. The experimental results show that theproposed method has a very good mapping result on lack of case or no example of ontologyfor mapping, and can get more complex relationship mapping.Secondly, as the demand of semantic interoperability is growing, in order to satisfy moresemantic application, a huge, complicated structure of ontology has come out. The traditionalmapping technology in the treatment of mapping among light ontology shows very goodperformance, but mapping quality and mapping efficiency are not ideal to mappingtechnology which contain more entity relationship and more complex large-scale ontology. Todeal with mapping finding tasks between large scale ontology has become a current researchfocus. One way is to avoid comparison between two ontology of all entities. Choose mappingentity pair, which is likely to be the candidate for mapping pair, from two ontology, thisneeds the original candidate mapping set compression. In other words, set choose the moreaccurate entity from the original entity as candidate mapping set.This paper puts forward candidate set compression and mapping method based on the APclustering. This method will apply approximation transfer clustering thought in the ontologyof the entity of clustering, the original mapping candidate set compressed into mappingcandidate subset. Those which belong to the same kind of entity are for mapping thecandidate set, eliminate noise entity, and improve the mapping performance. In the semanticsimilarity calculation, the same body and different entity in the ontology adopt differentcalculation method, and takes into account the semantic similarity information and structuresimilarity information on cluster effect. For smaller ontology clustering, clustering results canbe directly output as m: n mapping and large-scale ontology produces clustering results whichcan be as a mapping candidate set output, and then do other targeted strategy mapping, inorder to generate more accurate mapping results.Besides, ontology mapping strategy development is mainly based on different ontologyof the entity in similarity calculation and these entities have various information types (E.g.the semantic information, structure information).These information can be understood as thecharacteristics of ontology, and single mapping method can not acquire ontology entity’s totalinformation. Therefore, many strategy applications can be extensively used by the currentmapping method.The paper proposes matching spatial multi strategy matching method, which usesmethod based on the string and semantics to construct strategy of matching space firstly, which depends on the analysis of strategies to determine whether to add matching space.This method mapped from different entity similarity which is calculated by strategies, andoutput the best matching condition. Then, in this state use the strategy iteration based on thestructure matching to map results. With the reasonable combination of various matchingstrategies, it can avoid the insufficiency that a single method can not use all the informationin the ontology, allow the user to select various matching strategy flexibly, and provide a goodframework for a variety of tactics combination, making the mapping result more feasible.Finally, the paper sets out the method for ontology mapping based on the theory ofdecision making, using a rapid method for ontology mapping to get the candidate mappingentity, and then preselecting the strategy according to the information theory analysis and itssimilarity information, later using entropy decision analysis method to combine the chosenstrategy, getting the final mapping results. This method avoids the affect of uselessmapping strategy on the mapping results, provides users with an effective way of selectionstrategy, and use the automatic adjusting threshold to improve mapping precision andrecall. By the end of the calculation of the chosen strategy, Analyze and adjust thecombination weights. Experimental results show that using the choice of strategy and theadjustment of parameters can improve the overall performance of the mapping.