Research on Software Engineering-based Semantic Network Construction Techniques for Language Repositories

With the development and popularization of Internet technology, society has put forward higher requirements for the form and development process of software. In this study, firstly, based on the O-RGPS metamodeling framework, which provides modeling perspectives and meta-relationships among them for software engineering domain modeling, the semantics of ontology, role, and target models in the application domain are constructed. An extended semantic web is utilized to make the semantic network correspond to sentence information. Then, in conjunction with the Bayesian Confidence Network, the established semantic elements are transformed into ontologies, roles, and goals in the semantic process model. Finally, the heterogeneous semantic network based on semantic representation is constructed by using the Summary and Speak expression functions combined with the semantic knowledge of the heterogeneous users to specialize the representation of the ontology semantics. A new disambiguation mechanism for word meanings is created using the maximum entropy method, and an application case test analysis is performed on this semantic network. The experimental results show that the point degree distribution (u =1.9317, σ =0.6083) and weighted point degree distribution (u =2.0538, σ =0.7908) of this semantic network conforms to the lognormal distribution. The lognormal distribution pattern is still present in the case of blocking semantics. It shows that the study constructs semantic networks independent of the semantic order, which provides an important foundation for agile software engineering development and meeting the goals of user personalization and diversity. © 2024 Yun Li et al., published by Sciendo.