OLAP is a popular technology to query scientific and statistical databases, but their success heavily depends on a proper design of the underlying multidimensional (MD) databases (i.e., based on the
fact / dimension
paradigm). Relevantly, different approaches to automatically identify
are nowadays available, but all MD design methods rely on discovering functional dependencies (FDs) to identify
. However, an unbound FD search generates a combinatorial explosion and accordingly, these methods produce MD schemas with too many dimensions whose meaning has not been analyzed in advance. On the contrary, i) we use the available ontological knowledge to drive the FD search and avoid the combinatorial explosion and ii) only propose dimensions of interest for analysts by performing a statistical study of data.