Topic Modeling for Requirements Engineering: An Analysis of Ridesharing App Reviews

Abstract

Research in AI technology has become more popular today, helped by the rising of data volumes, powerful algorithms, and easier access to high-performance computing. Natural Language Processing (NLP) as a subset of AI technology plays an important role in the future of conversational AI because of its capability to interpret our natural language. On the other hand, ridesharing app industry is growing exponentially, helped by the rise of mobile device technology and the need for faster and cheaper mobility options. In the current thesis, we provide an overview of the current industrial practices in the development of NLP applications for analyzing app reviews and identify the gap in the state-of-the-art practices. To bridge the gap, this thesis proposes a method to extract information from the app reviews, with the goal to help ridesharing app developers to identify which features are most needed and which are less important. The proposed method is compared with the other similar methods and is validated with Europe’s top 10 ridesharing apps, including Bolt, Uber, Blablacar, Cabify, Via, Getaround, OlaCabs, Taxi.eu, Freenow, and Yandex Go. This contribution helps the ridesharing app developers to determine the requirements for developing their apps.