Integrating an AI-Driven Chatbot for Natural Language Traffic Safety Analysis within the SAFE-AL Dashboard

Abstract

This thesis presents the design and implementation of an AI chatbot by integrating into the existing Safety Analytics For Evaluating Alabama (SAFE-AL) dashboard to give users another way to interact with and understand transportation data. The chatbot func tions as a natural language query interface, helping users interact with and navigate traffic safety datasets without requiring technical knowledge of dashboard filters, SQL queries, or geospatial tools. Through natural language inputs such as “Show me Jefferson County in January 2025?”, the chatbot interprets the user’s request through a combination of both natural language processing and rule-based parsing and large language model use to navigate, understand and inform the user on such topics as time, location, and roadway. The pulled filters are then reviewed and sent through dashboard’s existing callbacks module, stored proce dures, and user interface to produce results. Rather than creating SQL queries or directly accessing the database, the chatbot operates as a navigational and conversational tool for the SAFE-AL dashboard. The chatbot remains read-only and uses the existing dashboard logic instead of creating new queries or changing the transportation data. In addition to text responses, the chatbot can automatically zoom to selected counties, highlight queried results on the dashboard map, and open related charts leveraging the existing callback logic already present in the dashboard. The research uses quantitative evaluation methods. Quantitative analysis compares Google Gemini and OpenAI ChatGPT using filtering accuracy, consistency with dashboard results, intent and understanding accuracy, hallucinations, and response time. The study also compares chatbot response times with manual dashboard interactions and reviews the cost effectiveness of both models. The study analyzes whether conversational an alytics can improve how users interact with transportation safety data while maintaining reliability, ease of use, and performance.