Título: Uma Proposta para Categorização de Textos por uma Rede Neural
Autores: Rizzi, Claudia Brandelero; Valiati, João Francisco; Engel, Paulo Martins
Resumo: The objective of this work is to present the results of two experiments in which a Multilayer Perceptron neural network, trained with the Backpropagation algorithm was used for categorization of texts in English (the categorization of texts is the classification of texts with regard to a group of one or more existent categories). The approach of organization of the input data adopted for this purpose is presented. It proved to be efficient in the training process, implying in a reduction of up to 1/3 the necessary time for its conclusion. The obtained results are satisfactory, since that in the first experiment, accomplished with the “IA Collection”, 72% of recall and 74% of precision were reached, and in the second one, accomplished with a sub-collection of Reuters-21578, 79% of recall and precision were obtained.
Código DOI: 10.21528/CBRN2001-109
Artigo em pdf: 5cbrn_109.pdf
Arquivo BibTex: 5cbrn_109.bib