Enhancing transparency and user-interactivity in sentiment analysis design through X-OODM

Abstract

Explainability is a significant factor in the realm of web-based applications. It provides a robust system for understanding and interpreting the internal functioning of applications throughout the design process. Nonfunctional parameters are integrated into the transparent and user-interactive models presented in X-OODM. Various components are employed to generate the metrics for each parameter, which then serve to develop the overall model metric. However, X-OODM used different scenarios of web-based applications as a case study to assess design quality metrics. In this study, we used domain diagrams from C0 to C10 as design models, improved with various sentiment analysis use cases, to assess the applicability of X-OODM and related metrics. Each domain diagram presents a distinct functionality that is evaluated under the user-interactive model and the transparent model of X-OODM. The user-interactive model uses transferability, informativeness, and accessibility, whereas the transparent model includes simulatability, decomposability, and algorithmic transparency. These parameters are further classified into several components, all of which contribute to the explainable model. A multiple linear regression is used to assess the explainable metric for each class domain model. The robust user-interactive and transparent model metrics determine the statistical significance for the design of web-based applications, specifically in sentiment analysis. This work can be extended to implement all the X-OODM models for the evaluation of web-based applications.