|Author||Anna Braasch, Bolette S. Pedersen|
|Title||Encoding Attitude and Connotation in wordnets|
|Abstract||The Danish wordnet, DanNet, though part of the global WordNet family, contains some information types that are not generally provided in wordnets such as qualia roles and connotation of words. Connotation is seen as the set of associations implied by a lexeme in addition to its primary, literary meaning; it is evoked by one (or more) particular feature of the entity referred to and suggests attitudes, emotions and opinions like admiration or disapproval. Thus, lexemes with a connotation have an observable pragmatic effect in texts making them subjective or opinionated.
In the paper, we discuss the relevance of connotation information in lexicons for computational applications in general and present the set of encoded semantic information exemplified by empirical data. We focus on a particular ontological type of entities, namely humans with the focus on selected hyponyms of person that are encoded with a connotation value and discuss the prototypical properties evoking positive or negative connotations. The qualia structure based approach enables to encode both the prevalent, connotation evoking features and prototypical activities of the person.
The material encoded with connotation so far consist of 650 nouns and comprises a male, a female and a gender-neutral group, thus it lends itself to comparative examinations concerning the distribution of connotation evoking features and polarity distribution within each individual group and between the groups as well. One of the most striking observations says that (in our material) the negative connotation polarity is predominant; the most important feature of female persons seems to be their positive appearance, and a general disparaging attitude dominates as regards the conduct and manners of male persons.
|Session||Computational Lexicography and Lexicology|