The study of the division of labour is undoubtedly one of the major topics in the history of labour and industrialization. In general, it has been addressed through the quantitative analysis of occupational categories in censuses, lists of professions or accounting records. This paper proposes an alternative approach, following new studies linking social history, digital humanities and tools to process natural language, and using a semi-massive corpus of primary qualitative sources to deal with this issue. We present how historians, computer scientists and NLP experts work together on this topic within the frame of the TIME-US research project, focusing, on the one side, on the methodology set to implement distant reading and, on the other side, on the resulting tools developped to support historians' research hypotheses.