Historical research regularly involves the intensive exploitation of serial and abundant bodies of data. For each assertion and each piece of data extracted, it is crucial to be able to go back to the source, and thus offer proof to the analyses carried out. Such corpora and questions can benefit from automatic textual analysis. Recent advances in artificial intelligence and automatic language processing, particularly in the field of large language models and generative artificial intelligence, enable us to process discourse over a long period and on a large scale. We aim to apply these methods to the parliamentary debates of the Third Republic (1870-1940). This corpus constitutes an exhaustive archive of legislative discussions, and automatic natural language processing (NLP) methods can help us extract relevant information, facilitate the analysis of political and social data, and compare and contextualize political speeches with historical events.