Automated Scoring System for Assessing Swiss Municipalities Sustainability

Diplôme Master Automated Scoring System for Assessing Swiss Municipalities Sustainability eng 2024 2025 author Arshad Muhammad Azeem 0009-0006-8310-5757 905787 director Dupuis Alexandre 100785 Département d'informatique (DINFO) 63 This thesis presents a research project to develop and implement an automated scoring system to assess the ESG factors across Swiss municipalities. Initially designed as a framework applicable to all municipalities in Switzerland, we apply it to certain medium-sized cities: Nyon, Rolle, and Vevey. By fine-tuning large language models based on the transformers architecture, we classified municipal council transcripts into ESG categories and subsequently applied sentiment analysis to derive ESG ratings. Our methodology included creating a new balanced dataset translated into French from a collection of English article headlines and fine-tuning CamemBERT models. Multiple models were produced using HPC, and a weighted voting scheme was employed to combine and enhance classification accuracy. The results indicated strong performance in identifying environmental, governance, and non-ESG-related content, with some challenges in distinguishing social aspects. Key findings revealed a significant representation of environmental topics and a notable increase in governance discussions. The social category may be underrepresented, possibly due to overlaps with the non-ESG category. Emerging trends showed stable ratings for Nyon and Vevey, while Rolle exhibited slightly more variability due to the lower number of available council transcripts. The automated system demonstrated efficiency in analysing large volumes of municipal transcripts, providing valuable insights for policymakers. Limitations included occasional misclassification of social content and a focus on French-language data. By translating the dataset into German or Italian, the framework could be broadened to encompass the entire country, and beyond. This work offers a novel approach for automatic ESG assessment in the public sector in Switzerland, facilitating more informed decision-making and setting a benchmark for future research in automated ESG classification. ESG NLP Environmental Social Governance Fine-tuning Automatic assessment CamemBERT Data engineering Rating ESG Rating Dewey 025.063 Master en sciences informatiques 32 Département d'informatique 18 false https://drive.google.com/file/d/1fu6KzpTnaAk3fDhedEafs5rXZrvMRhUT/view?usp=sharing https://drive.google.com/file/d/1bN_dKG9nHLLd7a9DAEmKwu1QegLFf0qr/view?usp=sharing 4572ccc5-0f05-4057-8adb-e43633f58a56 Master thesis Restricted MSc_Thesis_Report_AzeemArshad.pdf 3459259 application/pdf 68c0854a7a2a836c23bebb1e088470d302dac81e0821cc52336cd32243418aa4 restrictedAccess Faculté des sciences / Département d'informatique

Title	Published in	Access level	OA Policy	Year	Views	Downloads

Automated Scoring System for Assessing Swiss Municipalities Sustainability				2025	56	2

Archive ouverte UNIGE

Arshad, Muhammad Azeem

account_balance Affiliation entities

Affiliation entities