CN-VEFD: A visual emotion feature dataset for Chinese ESG reports

Area of Science:

• Environmental, Social, and Governance (ESG) reporting
• Corporate Finance
• Computer Vision

Background:

• Stakeholder perceptions are influenced by visual elements in ESG reports, yet research often neglects visual emotion.
• Existing studies primarily focus on textual analysis, leaving a gap in understanding the impact of visual cues.

Purpose of the Study:

• To introduce CN-VEFD, a comprehensive dataset of visual emotion features from Chinese ESG reports.
• To enable systematic research into the relationship between visual emotional signals in ESG disclosures and corporate governance.
• To facilitate exploration of how visual elements affect stakeholder responses.

Main Methods:

• Developed CN-VEFD from 13,481 ESG reports (2006-2023) of Chinese listed companies.
• Extracted 59 structured visual emotion features from 399,321 images, covering color, composition, and facial emotions.
• Quantified features using an automated pipeline integrating color analysis, saliency detection, and facial emotion recognition.

Main Results:

• Created a large-scale dataset (CN-VEFD) with 66 fields, linking 399,321 images to ESG reports and firms.
• The dataset captures visual emotion across color, composition, and facial dimensions.
• Features are systematically quantified for broad applicability.

Conclusions:

• CN-VEFD provides a valuable resource for visual analytics, emotion recognition, and behavioral finance research.
• The dataset supports the investigation of visual emotional signals in ESG disclosures.
• Enables deeper understanding of the interplay between corporate visual communication and stakeholder engagement.