Asymmetric Distributional Similarity Measures to Recognize Textual Entailment by Generality. (Mesures de similarité distributionnelle asymétrique pour la détection de l'implication textuelle par généralité)

Sebastião Pais

Abstract: Textual Entailment aims at capturing major semantic inference needs across applications in Natural Language Processing. Since 2005, in the Textual Entailment recognition (RTE) task, systems are asked to automatically judge whether the meaning of a portion of text, the Text - T, entails the meaning of another text, the Hypothesis - H. This thesis we focus a particular case of entailment, entailment by generality. For us, there are various types of implication, we introduce the paradigm of Textual Entailment by Generality, which can be defined as the entailment from a specific sentence towards a more general sentence, in this context, the Text T entailment Hypothesis H, because H is more general than T. We propose methods unsupervised language-independent for Recognizing Textual Entailment by Generality, for this we present an Informative Asymmetric Measure called the Simplified Asymmetric InfoSimba, which we combine with different asymmetric association measures to recognizingthe specific case of Textual Entailment by Generality.This thesis, we introduce the new concept of implication, implications by generality, in consequence, the new concept of recognition implications by generality, a new direction of research in Natural Language Processing.