Natural Language Generation from Concept Blends

Authors

Abstract

Conceptual Blending (CB) is a framework for the integration of knowledge from different sources onto one,
known as the blend. For example, the concept ?edutainment? is the blend of ?education? and ?entertainment?.
This paper presents an approach on producing a natural language reading of the blends that are generated by
Divago, a system that applies a computational implementation of CB to transform a search space. Represented
through a semantic network (the concept map), the blend demands considerable effort in its interpretation.
Obtaining its Natural Language description raises several issues and is an interesting challenge for the field of
Natural Language Generation. In this paper, we propose a reading based on diagnostic features with respect
to a reference concept. We make some reflections on the choice of these references, namely regarding the
creative potential associated.

Keywords

Natural Language Generation

Subject

Conceptual Blending

Conference

AISB'03 Symposium on Creativity in Arts and Sciences, April 2003