Construction and Analysis of Impression Caption Dataset for Environmental Sounds

TL;DR

This paper presents a new dataset of 3,600 impression captions for environmental sounds, generated with ChatGPT and validated by humans, to better capture human impressions associated with sounds.

Contribution

The study introduces a novel dataset of impression captions for environmental sounds, created using AI and human validation, filling a gap in existing sound-text datasets.

Findings

AI-generated captions are generally appropriate for describing environmental sounds.

Subjective and objective evaluations confirm the quality of the impression captions.

The dataset enables improved understanding of human impressions of environmental sounds.

Abstract

Some datasets with the described content and order of occurrence of sounds have been released for conversion between environmental sound and text. However, there are very few texts that include information on the impressions humans feel, such as "sharp" and "gorgeous," when they hear environmental sounds. In this study, we constructed a dataset with impression captions for environmental sounds that describe the impressions humans have when hearing these sounds. We used ChatGPT to generate impression captions and selected the most appropriate captions for sound by humans. Our dataset consists of 3,600 impression captions for environmental sounds. To evaluate the appropriateness of impression captions for environmental sounds, we conducted subjective and objective evaluations. From our evaluation results, we indicate that appropriate impression captions for environmental sounds can be generated.