Capturing algorithms in action. The case of the content qualification infrastructure of the Dailymotion platform

Publications on digital platforms are based on socio-technical infrastructures of qualification, the functioning of which remains largely unknown even though these platforms are extensively studied in the social sciences. Depending on the actors involved within an organization, the management, representation and final recommendation of content vary considerably. Highlighting the parts that make up the Dailymotion system, this article looks at the stages in which videos go from being complex assemblies of images, text and sound, to being lines in a database. It takes the TagMe text indexing algorithm as the point of entry to show how the qualification of content is both chosen and imposed on the organization. The initial orientations of its creators, the computational ontology adopted, the requirements of industrial video processing, and online advertising standards are all factors that define the configuration of the system and, for a time, guide the qualification of the videos.