Evaluating authenticity and quality of image captions via sentiment and semantic analyses

TL;DR

This paper introduces a method to evaluate the quality of image captions by analyzing sentiment and semantic richness, revealing insights into caption diversity and sentiment influence across a large dataset.

Contribution

It presents a novel evaluation approach using pre-trained models to assess sentiment and semantic variability in image captions, aiding quality control of crowd-sourced data.

Findings

Most captions were neutral with low semantic variability.

Approximately 6% of captions showed strong sentiment influenced by object categories.

Generated captions had minimal strong sentiment, unaffected by object categories.

Abstract

The growth of deep learning (DL) relies heavily on huge amounts of labelled data for tasks such as natural language processing and computer vision. Specifically, in image-to-text or image-to-image pipelines, opinion (sentiment) may be inadvertently learned by a model from human-generated image captions. Additionally, learning may be affected by the variety and diversity of the provided captions. While labelling large datasets has largely relied on crowd-sourcing or data-worker pools, evaluating the quality of such training data is crucial. This study proposes an evaluation method focused on sentiment and semantic richness. That method was applied to the COCO-MS dataset, comprising approximately 150K images with segmented objects and corresponding crowd-sourced captions. We employed pre-trained models (Twitter-RoBERTa-base and BERT-base) to extract sentiment scores and variability of semantic embeddings from captions. The relation of the sentiment score and semantic variability with object categories was examined using multiple linear regression. Results indicate that while most captions were neutral, about 6% of the captions exhibited strong sentiment influenced by specific object categories. Semantic variability of within-image captions remained low and uncorrelated with object categories. Model-generated captions showed less than 1.5% of strong sentiment which was not influenced by object categories and did not correlate with the sentiment of the respective human-generated captions. This research demonstrates an approach to assess the quality of crowd- or worker-sourced captions informed by image content.