Article Preview
Top1. Introduction
The world is changing with speed of light so new technology has come in market for data storage, data processing and data analysis. New technologies including streaming data, data from connected devices on internet of things, cloud computing, social media, high tech power grid, is driving a much greater volume of data (Panwar and Bhatnagar 2020). So, in the context of the Semantic Web Services, semantic interoperability based on ontologies has become an important challenge. Today, several strategies are proposed to improve this interoperability and solve heterogeneity problems ontological, such as strategies of alignment, integration, fusion and articulation.
In the semantic web infrastructure that rests on the description logics for the ontology construction, the alignment of the ontologies is based on the insert of the semantic entities of the local ontologies to the global ontology that respects the notions of subsumption and classification. In this case, the author must identify the existing semantic relations (equivalence, subsumption, disjunction …) among these entities which are already organized by the subsumption relation in the local ontologies (Kolli and Boufaida, 2009). The construction of the global ontology corresponds to the application of a succession of operations of change (add new mappings). This is a critical task, because the new implementation of change can make the global ontology incoherent. Inevitably, some mappings will be a source of contradictions in terms of ontology. According to (Kolli 2016), the contradictions in the ontology alignment can occur for several reasons such as:
- 1.
Mapping redundancy: Redundant mapping is one of the most common bugs.
- 2.
Mapping inconsistency: This problem is caused when two concepts have two different semantic relations by two different mappings or when two concepts change their relationship in the locale ontology to another kind of relation after mapping them.
- 3.
Mapping circle: Some suspicious mappings do not belong to the above two categories. It is difficult to find them in ordinary ways. However, such kind of mappings often shows abnormal behavior. Namely, if two entities are close (such as siblings) in an ontology, but they are mapped to two concepts of another ontology, which are far away from each other (Wang and Xu, 2008).
However, the majority of the ontology matching approaches like: GLUE (Doan et al. 2002), RiMOM (Tang et al. 2006) and VBOM (Eidoon et al. 2007) are only based on the measures of similarity between the semantic entities for the ontology matching process. These measures have as result a value that determines the similar entities (only equivalence relation) of the compared ontologies. This means that these measures are not sufficient for the construction of the global ontology. Furthermore, they cannot solve all the equivalences problems among the semantic entities for example the authors can have the same value for two different calculations because they rely on syntactic and structural criteria. Evaluation studies have shown that existing approaches often trade off precision and recall. The resulting mapping either contains a fair amount of errors or only covers a small part of the ontologies involved (Euzenat et al. 2006) (Euzenat et al. 2007) (Caracciolo et al 2008). Recently, there are some works done on handling incoherencies in distributed ontologies connected by alignments, where an alignment between two ontologies is a set of mappings between entities of the ontologies like (Euzenat, 2014), (Villazón-Terrazas et al., 2011) and (Euzenat, 2015).