Feature Selection and Negative Evidence in Automated Text Categorization [ Poster Paper ]

We ta kle two di erent problems of text ategorization, namely feature sele tion (FS) and lassi er indu tion. We propose a new FS te hnique, based on a simpli ed version of the 2 statisti s and a novel variant, based on the exploitation of negative eviden e, of the well-known k-NN method. We report the results of systemati experimentation of these two methods performed on the Reuters-21578 ben hmark.