Type inference on wikipedia list pages

The extraction of information from Wikipedia has led to a huge amount of knowledge made widely available by projects like the DBpedia. So far, most effort is put into extracting explicitly encoded information e.g. infoboxes. However, Wikipedia also contains a huge amount of implicit knowledge. One example for an untouched source of implicit knowledge are Wikipedia’s List of pages, in which multiple entities with a common type are collected. If this common type is known, it can be added to all entities of the list. Moreover, entities which are part of this list but not yet presented in the DBpedia can be added. This offers a huge potential for extending the DBpedia by adding missing type information. This paper proposes an approach to extract the shared types of a list using statistical methods and natural language processing. For a list entity, it was possible to infer new types with a precision of 86%.