Different versions of the same dataset show that concepts change their meaning over time, especially if the time gap is wide. Although not meaning exactly the same, two time gapped instances of the same concept may preserve some degree of sameness. For example, the concept shoemaker in the 17th century (someone who makes shoes with leather) has drifted until nowadays (someone who owns a company). Mapping drifted concepts correctly is necessary to solve longitudinal queries in Humanities data. [...] We will study what precise relationship holds between two different versions of a changing concept, identifying the presence of a drift and its nature. Using Description Logics work on ontology diff [3], we will define a minimum meaning concept core, which keeps stable over time despite other non essential transformations. A data model to represent drifted concepts will be needed. A systematic comparison between unstable concept properties will tell whether a drift occurred, and its type.

Humanities ontologies require dynamic concept formalizations instead of static ones, especially for contested, open-textured or ambiguous concepts. The definition of such concepts needs to be dynamically built depending on their contexts. Examples of contextual knowledge are the time when and the space where the concept occurs, subjective opinions on the concept, or domain expert statements about the concept. Multiple contradictory definitions may need to coexist in one ontology.

Concept drift and knowledge from contexts are closely related. Since the con- text of a concept often changes over time, a definition of concept drift based on the varying properties in contexts can be established.