Assessing Universal Accessibility of built environment though sentiment analysis at post-industrial disaster site – A case of UCIL, Bhopal, Madhya Pradesh

Abstract. In 1969, Union Carbide India Limited (UCIL) established a pesticide plant in Bhopal, India. On December 3, 1984, a massive leak of over 40 tons of Methyl Isocyanate gas from the plant turned the city into a toxic gas chamber, causing one of the world’s worst industrial disasters. An estimated 15,000 to 30,000 people lost their lives, and over 100,000 continue to suffer from chronic health issues. Decades later, the area around the UCIL site remains marked by environmental degradation, urban decay, and marginalization. This study focuses on assessing the universal accessibility of the built environment in the disaster-impacted zone using Sentiment Analysis. A study area was delineated around the UCIL site based on defined criteria and twenty-five landmarks were identified with study area. Google reviews were extracted for those landmarks and then filtered according to predefined universal accessibility key words. Selenium WebDriver, a robust browser automation tool, was employed to scrape reviews programmatically. After successfully collecting review data in CSV format, the next step was to analyze the sentiment of these reviews, particularly with respect to universal accessibility. For that purpose, the state-of-the-art Llama 7B natural language processing (NLP) model was utilized. The Llama model assess the sentiment of the reviews concerning accessibility features of the location. The model was instructed to assign a score between 0 and 1 for each review. The score of 0-0.4 represented the negative sentiment, while 0.5- 1 indicated the positive sentiment. The findings revealed that overall sentiment was negative in the study area with sentiment score not exceeding 0.23 for any landmark. The outcomes of sentiment analysis were cross validated using with GIS map. The results confirm the efficiency of the proposed method of Sentiment Analysis in Artificial Intelligence environment and need for such methods.