Thiago Salles,  Leonardo Rocha,  Fernando Mourão,  Gisele L. Pappa,  Lucas Cunha,  Marcos André Gonçalves,  Wagner Meira Jr.. 

In this work, we present temporally robust technics for Automatic Document Classification (ADC). This robustness is important since, due to language and knowledge dynamics, the characteristics of documents vary over time, causing a negative impact on ADC. We defined a temporal adjustment factor and, with it, we derived temporally robust classifiers based on traditional classification methods (Rocchio and KNN). Experimental results, obtained using two real textual datasets point to significant gains up to 11% compared to the traditional versions, and up to 4% compared to SVM (at a significantly smaller time).

http://www.lbd.dcc.ufmg.br:8080/colecoes/sbbd/2009/008.pdf

Processando consulta.