Categorización de texto en bases documentales a partir de modelos computacionales livianos

Abstract

We introduce a new text categorization method for documentary databases. The proposed method is an extension of the Naive Bayes text categorization model which allows obtaining good performance results in documentary databases with unbalanced training data. Experimental results allow us to conclude that the categorization method overcomes Naive Bayes and compares favorably with more sophisticated categorization methods such as support vector machines and logistic regression without increasing the use of computational resources in the training phase. [PUBLICATION ABSTRACT]