Opinion Mining to Aid User Acceptance Testing for Open Beta Versions

Abstract

Social media enables the sharing of opinions, ideas, interests, thoughts, hobbies, etc., by creating social platforms such as Twitter, Facebook, TripAdvisor, Yelp, Rooter, Goodreads, etc. The above sharing generates a lot of information over various social media platforms, which is used in different research areas. Among the many research areas, one is opinion mining, which identifies and extracts subjective information from source materials by using text analysis, natural language processing, and computational linguistics. The application of opinion mining is abundantly seen in the areas of marketing, governance, tourism industry, etc. However, the application of opinion mining is not much explored in the area of user acceptance testing. Therefore, in this research article, we propose a model for opinion mining and use it as a supporting tool for user acceptance testing. Specifically, we use the opinion mining on “Twitter” tweets to analyze the response of an Open Beta, i.e., Public Beta versions of the software to quantify acceptance criteria qualities. Hence, we try to inculcate opinion mining in earlier stages of software development to aid user acceptance testing in order to get the user feedback as accurately as possible to make change in the concerned software before its final version is released. The model is implemented and evaluated for Open Beta versions of IOS 13 and AndroidQ using machine learning techniques such as decision tree, K-nearest neighbor, multilayer perceptron, and logistic regression. Our evaluation shows that we achieved the best result with K-nearest neighbor having an accuracy of 97.12%, which is 14.37% higher as compared to the highest accuracy ever obtained in this area.