A natural language processing framework for analyzing public transportation user satisfaction: a case study

Abstract

Public transportation services make an important contribution to the nation's economy. However, the public transportation system was significantly impacted both during and after the Covid-19 outbreak. To minimize these impacts, it is important to know the users' sentiment and improve the service quality accordingly to change the users' attitude towards public transportation systems. Natural language processing is used to make meaningful inferences about user sentiment using various analysis techniques. Historically, surveys have also been used for years to learn users' opinions about transportation services. In this study, this traditional method was used to determine the satisfaction of public transportation users. The categorization model employed in the system developed as part of this work is based on algorithms such as Long Short-Term Memory (LSTM), Random Forest (RF), and Multi Logistic Regression (MLR). The dataset contains information gathered from the online survey. Of the models created utilizing the training dataset, it was discovered that the LSTM model offered the highest accuracy. Users' comments can help improve public transportation operators' operations, improve service quality, and monitor actions accordingly. Therefore, in this study, users' emotions were classified as positive, negative, or neutral based on the comments.