Augmenting app reviews with app changelogs: An approach for app reviews classification

Abstract

Recent research on the automatic classification of app reviews either focused on grouping app reviews into categories relevant to software evolution, or employed app reviews as the only research data to improve app reviews classification. Although it was reported that app review classification can benefit from supplementing user reviews with the data from other sources, only a few studies employed app changelogs for this purpose. This paper explores how to augment app reviews with changelogs to improve the accuracy and performance of classifying functional and non-functional requirements in app reviews. Specifically, we propose AUG-AC as an approach to extract feature words from app changelogs and construct the augments for app reviews. Next, we designed a series of experiments to evaluate our approach, varying in the length of AC-based augments for app reviews. The results show that AUG-AC outperforms the existing method by using app changelogs as a source of data next to app reviews.