Studies in Informatics and Control – ICI Bucharest
Sengodan MANI1*, Samukutty ANNADURAI2
1 Assistant Professor, Department of Computer Science and Engineering,
Nehru Institute of Engineering and Technology, Coimbatore-641 105, Tamil Nadu, India
maniscovai@gmail.com (*Corresponding Author)
2 Advisor & Senior Professor, Department of Computer Science and Engineering,
Hindusthan College of Engineering and Technology, Coimbatore-641 032, Tamil Nadu, India
anna_prof@hindusthan.net
Abstract: The web of data in the field of information and communication technology has been growing steadily in recent years, but there is a severe lack of association between the similar domains. The published web of data often reflects identical types of data that are described in various formats and generated at different locations. Interoperability and heterogeneity problems are generated when accessing such data, which can be solved by combining related ontologies with similarity-matching techniques. The semantic web allows information to be interpreted more meaningfully, providing a description of its contents and services in a machine-readable form called the web of data, which is typically structured in metadata. To process this data, different ontology matching methods are available, such as Silk, Kno Fuss, GLUE, LogMap, AgreementMaker, LIMES, CODI, SERIMI, RiMOM, etc. These approaches focus primarily on the classes of entities and their relationships, not on the principles of each type. In this paper a new model of similarity matching techniques is presented with the purpose of integrating related ontologies. The proposed model includes the entity behaviour and structural information of the ontology classes for the similarity matching process. The paper also includes two machine learning approaches, the first being lexical-based similarity matching using the Threshold-based Support Vector Machine (TSVM), whih is performed with restricted clustering and classification. The second is instance-based similarity matching using the Semantically enhanced Nearest Neighbour method (SeNN), which is employed in order to compare and quantify the semantic enhanced nearest neighbour entities/labels to predict the exact similarities. The final process involves the mapping of two sets of links based on the similarity of domain/ range criteria for accurate results. The proposed approach is compared with the existing state-of-the-art systems and the findings are analysed for precision, and specificity with respect to f-measure values that show better results in comparison with current approaches.
Keywords: Constrained based matching, Threshold-SVM, Semantic enhanced Nearest Neighbour (SeNN), Lexical similarity, Ontology matching, Similarity matching.
>>FULL TEXT: PDF
CITE THIS PAPER AS:
Sengodan MANI, Samukutty ANNADURAI, Explicit Link Discovery Scheme Optimized with Ontology Mapping using Improved Machine Learning Approach, Studies in Informatics and Control, ISSN 1220-1766, vol. 30(1), pp. 67-75, 2021. https://doi.org/10.24846/v30i1y202106