Article Preview
Top1. Introduction
Ontologies are formal representation of concepts with relationship between them within a domain, and they are represented as structured data models in the semantic web (Gruber, 1995). An ontology reflects knowledge from a certain domain of discourse. Ontologies naturally exist in an open web area, and their indirect (as an alignment) or direct (as imports) linkage is widespread (Zamazal, 2020). Deductive inference about classes and attributes is required to extract information from a data model. There are two primary issues, first is heterogeneity and second is interoperability that must be addressed during the information extraction process. Typically, ontologies are distributed across the web due to the distributed nature of the semantic web with possibility of similar data being found among them. Mapping between various ontologies can be used to solve heterogeneity. The majority of ontology mappings determine the similarity between the entities of different ontologies (Hussain, 2012). Entities can be used to describe concepts and their relationships. Several ontology mapping schemes have been suggested to quantify the similarities (Alasoud et al., 2009). All of them are mainly focusing on subject derivatives of the concerned domain.
As the number of ontologies increases, the need to communicate with one another become more complex . It is necessary to develop an interoperability mechanism in order to collect accurate data. Due to the distributive nature of the semantic web, ontology heterogeneity makes the interoperability process far more complicated. These scenario allows the development of effective and successful ontology matching approaches. There are several types of ontology matching approaches have been proposed by semantic web groups and individuals (Anan et, al., 2015; Somasundaram et al.,2019;(Anam et al., 2015) Ferranti et al.,2021). Many of the systems are mainly concerned with the subject derivatives of the problem domain. Further, numerous applications such as data integration, ontology evolution, data warehousing, e-commerce, and data exchange in diverse fields require ontology mapping (Anam et al., 2015). In this paper, an improved structural-based ontology matching approach that uses similarity spreading is proposed that provided with data models in a structured format. This suggested model addresses ontology mapping by analyzing the structural features of ontologies. The mapping technique of this approach essentially comprises node clustering based on edge affinity, followed by graph matching using coefficient similarity propagation on each clustering. Similarity score is calculated at the end of each iteration (Zager & Verghese, 2008) and the combined similarity aggregation is performed in cluster based ontology matching approach (Tran et al., 2011).
The propagation technique requires an initial seed which often a similarity pair derived through the terminological similarity method. The initial seed is propagated to each part of the node clustering set using similarity coefficient propagation method. The process is repeated for each node clustering set and the similarity score is measured and aggregated at the end. To achieve efficient similarity matching between different ontologies, similar pairs are found along the path by comparing the siblings of the nodes in the clustering (Laadhar et al., 2017). In terms of accuracy, recall and f-measure parameters, the proposed model outperforms similar systems.
The purpose of this research is to develop an efficient ontology-matching model with a wider set of ontological properties. The innovation entails a method of propagating coefficient similarity values along the path of graph that is limited solely to clustered nodes and eliminating the need for propagation on the subgraph.
The rest of the paper is organized as follows: section 2 reports the review of the literature, section 3 discusses the methodology for ontology alignment and illustrates the proposed ontology mapping system, section 4 discusses the results and evaluates the proposed system's performance, and section 5 concludes the novel method and suggests some guidelines for future ontology alignment work.