Generating Affective Captions using Concept And Syntax Transition Networks
T. Karayil, P. Blandfort, D. Borth, A. Dengel
Abstract
We present a method for generating affective image captions using Concept and Syntax Transition Networks, combining visual concept detection with structured language generation to produce emotionally expressive descriptions.
Overview
This work addresses the challenge of generating image captions that convey emotional or affective content. Traditional image captioning systems focus on factual accuracy, but human descriptions often carry sentiment and emotional nuance.
Concept and Syntax Transition Networks
We propose Concept and Syntax Transition Networks (CSTN) as a framework for generating captions that incorporate affective elements. The method combines visual concept detection with structured language generation to produce emotionally expressive descriptions of images.