Properties are extracted from 73795 recipes taken from Recipe Source. Ingredient properties are extracted both from the ingredient list, which includes some cooking actions, and from the textual preparation, from which the actions applied to the ingredients are extracted using NLP tools [8]. The latter approach uses both classical morpho-syntactic analysis methods and a discourse representation specifically devel- oped for procedural texts (instruction texts, including recipes). The discourse representation is well adapted for solving some complex anaphoric phenomena appearing in those types of texts. An example of a recipe used for property extraction is given in Fig. 4. Extracting the properties of the ingredient list is done through an automatic annotation process, because the recipes are initially in a textual form. This annotation