The 19th and 20th centuries were a period of dramatic changes in France, as the way people lived, worked, and interacted, and the places where they did it, changed beyond recognition. The general picture is well documented as historians, economists, and sociologists have studied urbanization, industrialization, and the demographic transition and their role in the long-term evolution of French society. Much less is known about the spatial variation of these phenomenon, their mechanisms, and their effects. Thus, there is a discrepancy between the complete national macro-data that gives the “big picture” of the evolution for France as a whole (or, at best, at the département level) and the various local micro-data that offer a much more precise history for some locations but are difficult to generalize. This gap not only prevents more fine-grained analyses of underlying dynamics of historical evolution, but it also limits our understanding of the mechanisms of social change. Relevant archival material exists, but the methods to collect and exploit it on a national scale are yet to be created. This is the aim of the SOCFACE project, which proposes technical innovations from a computer science perspective in order to build invaluable resources for the social sciences. SOCFACE will process, for the first time, all listes nominatives du recensement from 1836 to 1936. Produced every five year, these listes are organized spatially (municipality; wards, hamlets, or streets; houses; households) and summarize the information from the census, listing each individual with some of his or her characteristics, e.g., name, year of birth, or occupation. The sheer number of listes (around 15 million images from 1836 to 1936, corresponding to approximately 700 million individual records) and their dispersion over space (they are located in almost a hundred archives depository) limited their use until know. We will develop automated methods to extract the information they contain. Indeed, automatic recognition of handwritten documents does produce today usable transcription but it relies excessively on large samples of annotated data from the initial corpus or on preexisting linguistic models. Putting together archivists, historians, economists, demographers, and computer scientists, SOCFACE aims at overcoming the technological limitations that prevent automatic processing of a huge but homogenous collection of historical sources–with the same structure but written by tens of thousands of different people–by developing auto-supervised recognition models. In addition, and to better harvest the wealth of the listes, SOCFACE will link individuals between censuses using automated methods developed recently for the US census, adapting them to the French sources, relying on our long experience in building historical datasets and linking records on a large (albeit much smaller) scale. SOCFACE will be the first project in France to link records from the same individuals over time on a large scale, opening huge opportunities for research, both during the project and afterwards. Thus, SOCFACE will represent an unprecedented advance relative to existing methods to process extremely large series of handwritten historical documents, that could be used to decipher many other potential archival documents. SOCFACE will produce a complete database of all individuals who lived in France between 1836 and 1936 and use it to analyze social change from the ground up, greatly improving our understanding of French economic and social structures. The other major impact of SOCFACE will be for the access to listes nominatives by the general public. Thanks to FranceArchives, we will disseminate the information available in the listes in Open Access, allowing anyone to browse hundreds of millions of records freely. This will be a major improvement compared with the limited number of observations and recollection, mostly partial in time and location, available today.