Research On Ontology Mapping Based On Semantic Web

Semantic Web uses metadata with semantic information to annotateresources on the web so that machines can understand them。In philosophy, anontology is a theory about the nature of existence, of what types of things exist.Gruber presented the definition of ontology which is used commonly today:"An ontology is a formal, explicit specification of a shared conceptualization."The concept and method of ontology are used by computer research to realizeknowledge expressing, sharing, and reusing. Ontologies are cores in theSemantic Web because they are the carriers of the meaning contained in theSemantic Web. However in many cases, different domains define differentontologies containing the same concepts. Even in the same domain, differentorganizations construct different ontologies. Therefore, it is necessary to find aflexible, practical approach to establish semantic correspondences betweenontologies and implement the exchange of data annotated by differentontologies.In the earlier development of the web, correspondences betweenontologies can be constructed manually by hand. With the rapid increase ofdata described by ontologies, finding such mappings manually is tedious,error-prone, and clearly no longer possible, so semi-automatic or automaticmapping methods are needed.So far, many different approaches have been proposed with diverse rangeof mapping techniques. For example, an integrated ontology mappingapproach was proposed based on rules and quick ontology mapping putattention to the runtime of program. Anchor-PROMPT is a tool for ontologymerging and mapping. It contains a sophisticated prompt mechanism forpossible mapping entities. At the same time researchers from severaluniversities are working together to create an ontology mapping approachbased on information flow. The approach of semantic enrichment for ontologymapping exploit text categorization to automatically assign documents to theconcept in the ontology and use the documents to calculate the similaritiesbetween concepts in ontologies.In ontology mapping, it is common to compute semantic similaritiesbetween concepts in entities. To achieve this, dictionaries and thesauri areneeded, such as WordNet. WordNet is a large lexical resource combiningfeatures of dictionaries and thesauruses in a unique way that allows for a freshperspective on the semantics of nouns, verbs, and adjectives and offers newpossibilities for exploring the internal structure of the lexicon. WordNet is awidely used semantic network which is organized by synset. Each synset maycontain multiple words with similar meanings. The objects that WordNetdescribes include compound, phrasal verbs, collocations, idiomatic phrases,words and words are the most elementary units in it. Although WordNet isorganized by synsets, it is not easy to enumerate synsets. Between synsetsthere are some relationships, such as hyponymy and meronymy. The lexicaldatabase WordNet is particularly well suited for similarity measures, since itorganizes nouns and verbs into hierarchies of is–a relations. We use WordNetas auxiliary information to calculate similarity values between concepts in thetwo ontologies.In this paper, we use instances in ontologies to enrich the structure ofontologies and then use vector space model to deal with these instances. In thevector space model, if information capturing system refer n key words, we willconstruct n-dimension vector space and each dimension express different keywords. Before deal with these instances documents, we will make thempreprocessed and this process is just like a kind of reducing dimensionsapproach. There are many approaches to computing weights of words and wedeploy the method developed in Smart system. It considers both the frequencyof the word appearing in a document and the number of documents thatcontain the word. It guarantees that a word, which has a high appearancefrequency coupled with a low number of documents containing it, has a highweight. The vector space model make the evaluation of key words possibleand we can scale the similarity of entities in ontologies bu using these concretedata.In this paper, we proposed an ontology mapping approach ACAOMwhich combines two strategies. These two strategies make use of nameinformation and instance information assigned to concept nodes respectivelyto calculate similarities between entities. Then an integrated approach isdesigned to incorporate both strategies. The experimental results show thatACAOM performs better than iMapper and it improves the precision ofiMapper from +2.4% to 5.9%.There are several aspects that can be improved in our proposed system.1. We could realize ontology merging and integration in the same system.ACAOM can be applied to other aspects of ontology related issues, such as,queries based on distributed ontology.2. Our method can not support n:m mappings at present, which are useful inmany cases, we will extend our method to deal with these cases in the future induring complex mappings.