Textual Entailment is a research task in NLP that aims at determining whether a hypothesis is entailed by a text. Usually tackled by machine learning techniques employing features which represent similarity between texts, the recent availability of more training data presupposes that neural networks that are able to learn latent feature from data for generalized prediction could be employed.
The University of Luxembourg, together with University of Turin, has developed a Deep Learning approach for textual entailment for question answering on legal texts, called LegalBot, able to generalize well without excessive parameter optimization (Adebayo et al., 2017a). Further improvements and applications are currently under investigation.
In (Adebayo et al., 2017b), the approach has been tested on the COLIEE dataset, which involves the identification of an entailment relationship such that, given a question and relevant articles, it must be determined whether the relevant articles entail the question or not. Specifically, the task was to automatically give answers of the type `yes’ or `no’, where `yes’ means that the question is entailed by a text (and `no’ otherwise).