Analyzing Customer Feedback on Public Transit using Large Language Models

Transit riders’ feedback provided in ridership surveys, customer relationship management (CRM) channels, and, in more recent times, through social media, is key for transit agencies to better gauge the efficacy of their services and initiatives. Through our research partnership with the Washington Metropolitan Area Transit Authority, we developed MetRoBERTa - a transit-topic-aware large language model (LLM) capable of classifying open-ended text feedback to relevant transit-specific topics, providing new insights on previously unstructured customer opinion data, and the ability to use customer sentiment to evaluate the effects of policy changes on a granular scale. This tool enables transit agencies to be more responsive to public opinions, developing a better relationship with the communities they serve.