The Scientific and Technical Center for Building (CSTB) was asked to develop a tool to help identify materials potentially containing asbestos in buildings. CSTB has a set of building descriptions, but these descriptions do not mention the used marketed product but only the product class. In this seminar, we will present the two approaches guided by a domain ontology developed to predict the presence of asbestos in building elements. The first approach is based on two external resources (the INRS1 and Andeva2) containing temporal descriptions of the marketed products and their asbestos characteristics, generates a probability of the presence of asbestos depending on the product classes and the date of construction of the building. The second approach, called CRA-Miner, is inspired by methods from Inductive Logic Programming (ILP) to discover rules from the positive and negative examples described in the CSTB knowledge graph based on a semantic context, and a set of heuristics adapted to part-of relations, defined by the expert. We will present the results obtained on 2998 building descriptions and show that if the precision is better in the case of the first approach, the second allows to take more decisions.