CoMSS: Context based measure for semantic similarity between conceptual models

Artifacts are defined as the tools which act as building block for developing a software or project. Artifacts are of several kinds like UML and BPMN. The overall connotation of the document in SDLC is combination of sensible or significant words represented in the form of different artifacts. In this approach main focus is on the precision of the documents or artifacts developed by requirement analysts. This phase is purely personal opinion oriented, later on which may cause rejection of the software during acceptance testing. Our aim is to preserve the overall meaning of the document that's why we compare the class diagram (conceptual model) with BPMN so that the meaning of the requirements provided by the stakeholders should be preserved in both forms of artifacts. The word specificity and word semantics plays vital role in assessing the semantic similarity between the two artifacts. In this paper the frequency (word specificity) is calculated between the vectors with the help of TF-IDF and the semantic similarity between the vectors is calculated with the help of WordNet algorithms. The proposed similarity measures are evaluated in divergent context, the benchmark dataset i.e SemEval (Semantic Evaluation) 2012 Semantic Textual Similarity test set, competition organized in 2012. Three standard case studies are used to evaluate the semantic similarity between the artifacts named as Rambaugh's ATM Model (Rambaugh et al. 1991), EFP (Kurt, 1995), Course Registration (IBM Corp, 2004).