A computational framework for predicting software quality-in-use from software reviews

• Main Author: Atoum, Issa
• Format: Thesis
• Language: English
• Published: Universiti Malaysia Sarawak, (UNIMAS) 2015
• Subjects: T Technology (General)
• Online Access: http://ir.unimas.my/id/eprint/10763/1/Issa%20Atoum%20ft.pdf; http://ir.unimas.my/id/eprint/10763/

Software Quality-in-Use (QinU) lies in the eyes of its users. QinU has gained its importance in egovernment, mobile-based, and web applications. Currently, QinU is measured using either ISO standard (e.g. ISO 25010) or customized model approaches. These approaches tend to be incomplete and suffer from problems of user’s task definition sizing. Therefore, QinU measurement by these approaches has complexity resulting from quantifying QinU systematically. This thesis proposes a computational and novel QinU Framework (QinUF) to measure QinU competently consuming software reviews. The significance of the framework is that it combines the semantic analysis and sentiment analysis research areas. In semantic similarity area, we proposed a novel Weighted Sentence Similarity Measure (WSSM) and developed an algorithm to predict a review-sentence QinU topic (QinU characteristic or software aspect). In the sentiment analysis area, we proposed an algorithm to classify and aggregate software review-sentences into QinU topics. Experiments showed that the QinUF was able to predict software QinU topics on the fly, with high performance compared to selected topic prediction methods. Moreover, results of built use cases showed that employing minimal set of QinU features (properties) enable users to acquire software easily. As for future research, it is recommended to extend QinUF to support additional QinU characteristics, enhance sentiment orientation, specialize the framework to a certain software domain, and implement the framework in a large-scale system.