An NLP-Based Framework for Sentiment and Topic Analysis of Citizen Feedback on U.K. Government Mobile Applications

Abstract

With the rapid growth of digital technologies and machine learning methods in recent years, governments have the opportunity to analyse and collect citizens’ feedback, which is usually unstructured and large, to improve their public services and citizen engagement. This study investigates an application of Natural Language Processing (NLP) in public administration by mining citizens’ feedback from U.K. government mobile applications. The research focuses explicitly on three widely used applications, including HMRC, NHS, and GOV ID CHECK. By examining these reviews, the research aims to uncover the overall sentiment towards these applications and identify the primary issues and areas for improvement as expressed by users. The study employs advanced NLP techniques, including sentiment analysis and topic modelling, to process and analyse user reviews collected from the App Store and Play Store. The result of the study shows four significant points: 1) sentiment analysis based on TextBlob labelling process achieved an accuracy of around 80% for all the apps, 2) users had more positive sentiment towards the HMRC app and expressed satisfaction with HMRC and GOV ID apps based on topics from LDA model, 3) ID Verification and Login issue were among most common challenges of all the apps, and 4) NLP proved to be a powerful tool for government to obtain and analyse citizens’ feedback towards the services.