Comparación de Enfoques para Evaluar la Validación de Respuestas

The Validation of Answers has been seen recently as a classification problem able to introduce Machine Learning for improving Question Answering results. The unbalanced nature of collections has led to the use of measures based on precision and recall. However, Relative Operating Characteristic (ROC) analysis is preferred sometimes in similar classification tasks. In this article we compare both approaches according to their rationale, their stability with respect to the size of the collection, their discriminative power and their adequacy to the particularities of the Answer Validation task.