The idea of CaLiGraph is to exploit the knowledge about resources from categories and listpages in Wikipedia. First, we derive an ontology from Wikipedia using the graph formed of categories and listpages. To that end, we have to clean this graph from nodes that do not express classes (like the category London) and edges that do not express hierarchy relationships (like the subcategory-relationship of Songs and Song awards). Then we combine this graph with DBpedia's hierarchy in order to form a large ontology. Additionally, we enrich this ontology with restrictions learnt from the Cat2Ax approach. After having formed a large ontology, we use this information to extract new resources (that have not been specified in DBpedia) from Wikipedia listpages using machine learning.
This section will be updated with an in-depth explanation of the ideas as soon as the paper describing CaLiGraph is finished.