Experiments with a Differential Semantics Annotation for WordNet 3.0

Abstract

This article reports on the methodology and the development of a complementary information source for the meaning of the synsets of Princeton WordNet 3.0. This encoded information was built following the principles of the Osgoodian differential semantics theory and consists of numerical values which represent the scaling of the connotative meanings along the multiple dimensions defined by pairs of antonyms (factors). Depending on the selected factors, various facets of connotative meanings come under scrutiny and different types of textual subjective analysis may be conducted (opinion mining, sentiment analysis).