Sense Groups


Cornetto2.0-sense-groups are relations between senses of Cornetto2.0, according to the dump of Cornetto of the 8th of July, 2012. Modifications to Cornetto after that date are not reflected in the sense-groups.

We derived 4 sets of sense-groups for lemmas based on different relations. A sense-group is a set of meanings of a word that are semantically close and therefore difficult to discriminate both for humans and machines. Metonymy, specialization and generalization of meaning of words can lead to closely related meanings that are compatible and can apply simultaneously in a context. An example of metonymy is “academie” (academy) referring to the institution or the building. A case of specialization/generalization is “behandeling” (treatment), referring to a medical treatment but also to treatment in general.

Metaphorical meanings are considered not to be compatible: i.e. in a context both meanings cannot both be true or relevant. E.g. “slang” (snake) can refer to the animal, a person, a tube or a snake-like structure or form but never to combinations. Unrelated meanings are often coming from different origins and have the same spelling by accident, e.g. “pad” (toad/path), referring to an animal or a small road. Both metaphorical and unrelated senses are intended to be excluded from sense-groups because their meanings are not compatible and can more easily be distinguished.

By distinguishing sense-groups it is possible to apply WSD at different levels of precision and relevance.

The files can have overlapping lemmas, lexical units and groups.


Leave a Reply