Abstract

This thesis investigates how people in Montreal feel about transportation by analyzing posts on
X (formerly Twitter) using a Large Language Model (LLM) called GPT-4o. Montreal is a unique
city with French and English speakers, making public opinion mining challenging. However, GPT-
4o can directly process both languages, making the analysis more accurate and efficient.
Unlike traditional methods that often struggle to capture the nuances of language, GPT-4o generates
precise sentiment analysis, helping us understand the emotions behind people’s opinions. this
tool was used to categorize tweets into transportation modes (e.g., bus, metro, and train), specify
aspects (e.g., safety, cost, and punctuality), and overall sentiment (positive, negative, or neutral).
Local terms like ”REM” and ”STM” were included to ensure the AI understood the context. AIgenerated
aspects were then grouped into standardized categories like reliability, cost, safety, and
environmental impact to enhance clarity and consistency.
This approach showcases the flexibility and scalability of LLMs for multilingual public opinion
mining. The study revealed significant differences in public sentiment across transportation modes
and aspects, such as safety concerns in cycling and punctuality issues for public transit. These insights
are valuable for transportation planners and policymakers seeking to improve urban mobility.
Future research could explore other public opinion sources or use this technology for real-time
sentiment tracking to aid urban infrastructure planning.