Leveraging Topic Modeling and Sentiment Analysis to Improve Digital Bank Applications

Digital banking applications have experienced significant growth in the recent years. To create a positive perception towards the applications, customer reviews may be utilized to identify the possible improvements for the applications. In this research, firstly, the Latent Dirichlet Allocation (LDA) model is applied to obtain the topics that are discussed in the reviews. Five topics are discovered in the case study, in which the number of topics is selected by the coherence score metric. A sentiment analysis using SenticNet 5 lexicon is subsequently performed, along with the part-of-speech tagging and dependency tree parsing for each sentence, to obtain the sentiment towards specific attributes of the application. The Opportunity Landscape Map visualizes the connection between the topics and the sentiment, such that the user experience topic is identified as underserved in the case study. The examples of potential improvements are adding features for account mutation, paying credit card bills, and changing phone number.