Textual Entailment applied to Answer Validation

Supervisor: Dr.A.Peñas

This work in progress aims to improve the results of Question Answering (QA) systems, while promoting a better selfconfidence on their answers. For this purpose, an Answer Validation (AV) subsystem, able to decide about candidate answers correctness, has been developed and applied to a set of QA systems. This AV subsystem relies on Recognizing Textual Entailment (RTE) techniques, that take advantage of common paths in dependency trees to decide about the semantic similarity of texts. After building the proposed AV subsystem it was evaluated independently, obtaining promising results. Then, the set of QA systems including it as an extra
computing module was evaluated as well, showing the benefits of including such components as a part of QA technologies

Jesús Herrera de la Cruz