Following the Yellow Vest crisis that occurred in France in 2018, the French government launched the Grand Débat National, which gathered citizens’ contributions.This paper presents a semantic analysis of these contributions by segmenting them into sentences and identifying the topics addressed using clustering techniques. The study tests several combinations of French language models and community detection algorithms, aiming to identify the most effective pairing for grouping sentences based on thematic similarity. Performance is evaluated using the number of clusters generated and standard clustering metrics.Principal Component Analysis (PCA) is employed to assess the impact of dimensionality reduction on sentence embeddings and clustering quality. Cluster merging methods are also developed to reduce redundancy and improve the relevance of the identified topics.Finally, the results help refine semantic analysis and shed light on the main concerns expressed by citizens.