This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility: Every loaf of bread has a carbon footprint, a supply chain and a workforce behind it. As climate change, food safety and labor rights attract growing attention, corporate disclosures on environmental, social and governance (ESG) issues are becoming increasingly important.
But what do companies disclose, and have their priorities changed? A team at Kyushu University developed an AI-based method to assess how closely corporate disclosures align with the ESG issues that matter financially to each industry. Applying it to Japan's food and beverage sector, the team found that firms place greater emphasis on social topics, while environmental disclosure remains relatively limited.
Priorities also shifted as reporting requirements changed. The study was published in Corporate Social Responsibility and Environmental Management on Sept. 16, 2026. "High-quality ESG disclosure helps stakeholders assess business performance and risks," says Siyu Shen, a graduate student at Kyushu University's Graduate School of Economics and the paper's first author.
"Yet counting mentions, or judging whether the language sounds positive or negative, tells us little about how seriously a company treats those issues." Determining which issues matter most requires an industry-specific benchmark. The team turned to the Sustainability Accounting Standards Board (SASB), which identifies financially material sustainability issues by industry. The researchers focused on Japan's food and beverage industry, where reliance on agricultural raw materials and global supply chains makes ESG risks relevant to costs and stability.
They analyzed 1,623 annual securities reports filed by 340 publicly listed companies between 2020 and 2024. To handle that volume of data, the team used Text Match Pre-Trained Transformer (TMPT), a multilingual AI model designed for semantic text matching. TMPT first locates passages in each report that relate in meaning to ESG issues.
Rather than matching exact words, it captures concepts expressed in different ways, grouping words like "greenhouse gas (GHG) emissions" and "CO₂ emissions" under the same issue. The model then rates semantic relevance. A statement giving the exact tons of emissions a company cut, for instance, scores higher than a general pledge to work toward carbon neutrality.
Extract — continue reading at the source.
