The Influence of Semantics in Text Categorisation: A Comparative Study using the k Nearest Neighbours Method.
Edgardo Ferretti, Marcelo Luis Errecalde, Paolo Rosso · Indian International Conference on Artificial Intelligence · 2005
In this paper we investigate dieren t uses of semantics in text categorisation tasks. At this end, we consider distinct representa- tions of documents which dier in the kind of information incorporated: a) information about terms only, b) semantic information (terms sense) and c) a combination of both types of information. Moreover, we study how the vocabulary size reduction aects this task. Thek Nearest Neigh- bours method was used to perform the categorisation and the vocabulary size was reduced by means of the Information Gain technique. A num- ber of dieren t document codications were tested. The experimental results showed that in corpora richer syntactically and semantically the inclusion of semantic information improves the text categorisation task if vocabularies with a sucien t number of features are considered.