Authors:
Elena Cabrio
1
and
Serena Villata
2
Affiliations:
1
University of Nice Sophia Antipolis, France
;
2
CNRS, France
Keyword(s):
Argumentation, Textual Entailment, Natural Language Processing.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Artificial Intelligence
;
Knowledge Engineering and Ontology Development
;
Knowledge Representation and Reasoning
;
Knowledge-Based Systems
;
Natural Language Processing
;
Pattern Recognition
;
Symbolic Systems
Abstract:
Textual Entailment (TE) systems aim at recognizing the relations of entailment or non entailment holding between two text fragments (i.e. a pair). The identified TE pairs are considered as independent one from the others. However, in the latest years TE systems have been challenged against a number of real world application scenarios like analyzing costumers interactions about a service, or analyzing online debates. These applications have underlined the need to move from TE pairs to TE graphs where pairs are no more independent. Moving from single pairs to graphs has the advantage of providing an overall view of the topic discussed in the text. The challenge here is to define ways to exploit such graph-based representation for text exploration. In the literature, some approaches apply abstract argumentation theory to compute the accepted arguments of a debate, but they present a number of drawbacks, e.g., the non entailment relation and the attack relation in abstract argumentation
are assumed to be equivalent, but this is not always the case. In this paper,
we define bipolar entailment graphs, i.e., graphs whose nodes are text fragments and the edges represent the entailment or non entailment relations. We adopt abstract dialectical frameworks to define acceptance conditions for the nodes such that the resulting framework returns us relevant information for our text exploration task. Experimental evaluation shows the feasibility of our approach.
(More)