In this long paper, we use NLP techniques to explore two decades (1881-1899) of parliamentary debates of the French Third Republic (1870-1940), and more specifically to analyse the importance of the army in the political debate. We use Latent Dirichlet Allocation to partition the vocabulary into topics, and then study the distribution of the topic "army" over time. We also examine its connection with other topics, in relation to the main political and military events of the period.