This article is a Japanese-language commentary on a report co-authored by Managi, President of aiESG, Inc. and Senior Professor at Kyushu University. A summary of this report is included in the latest issue of the Asian Bond Monitor, published this month by the Asian Development Bank.ASIA BOND MONITOR JUNE 2025 Emerging East Asian Local Currency Bond Markets: A Regional UpdateThe information is published in the

summary

The Asian Development Bank is an international development finance institution established to support economic growth in Asia and the Pacific and contribute to the economic development of developing member countries. We are pleased to present here Unveiled by Deep Learning: The Environmental, Social, and Governance (ESG) Emphasis of Leading Companies in East Asia and Southeast Asia. Emphasis of Leading Companies in East Asia and Southeast Asia.

The report provides an in-depth analysis of how leading companies in eight East and Southeast Asian countries address environmental, social, and governance (ESG) issues in their corporate reports. Advanced natural language processing (NLP) and artificial intelligence (AI) tools, particularly "Text Match Pre-Trained Transformer" (TMPT)modelto bridge the gap in ESG reporting in multilingual and multicultural environments and to "visualize" a company's ESG focus from large unstructured text data.

The study included 480 annual and integrated reports collected from 293 companies with the largest market capitalizations in the securities markets of the People's Republic of China (PRC), Japan, Korea, Indonesia, Malaysia, the Philippines, Singapore, and Thailand. The analysis found that, overall, companies were more interested in economic and governance topics than in environmental and social topics, especiallyeconomic ripple effect, ,production cost, ,Governance RiskIt is clear that the emphasis is on

Background and Purpose of the Report

In recent years, investors have increasingly focused on information related to environmental sustainability, social responsibility, and quality of governance in evaluating corporate performance.ESG Factorshas become an important component of corporate strategy and reporting, but understanding the emphasis companies place on these topics, especially in a multilingual and multicultural environment, has remained limited.

According to the European Commission (2022), despite the fact that many market participants use ESG assessments, the ESG assessment market is not functioning well, and concerns have been raised about the lack of transparency and bias in assessment methodologies and low correlation between various ESG assessments.

With that background in mind, this report,Gaps in ESG ReportingThe purpose of this study is to address the In light of Yoshida et al. (2024) questioning the consistency and reliability of traditional ESG assessments in previous studies, this study provides a nuanced approach to assessing ESG-related discussions without imposing a rigid, uniform scoring system. In doing so, a key objective of this report is to provide a deeper understanding of the ESG strategies and disclosure practices of a diverse range of companies.

Analysis Details

At the heart of this study are,TMPTThe AI and deep learning model called

1. Data Collection and Scope

• Target Companies/Regions:. The 293 listed companies with the largest market capitalization in the securities markets of the People's Republic of China (PRC), Japan, Korea, Indonesia, Malaysia, the Philippines, Singapore, and Thailand were included.
• Report:. 480 annual and consolidated reports were collected, including annual reports issued in 2024.
• Languages: English Reports were collected in English and in each local language (where available).
• ESG Topics:. Human rights, labor environment, governance risks, greenhouse gas emissions, etc,13 ESG-related topicswere included in the analysis.
  

2. How the TMPT model works

• Training data: (1) TMPT is trained on a vast array of multilingual texts from Wikipedia and academic research.
• Analytical Processes:. Each corporate report is matched against a list of 13 ESG topics, such as climate change, human rights, and job creation. The report is broken down into smaller sections to see how well each section discusses these issues, and each company is given a score indicating its emphasis on each ESG topic.
• Features: -Features: -Features The model provides a scalable, multilingual approach to detecting ESG content without relying on predefined taxonomies, allowing for more consistent comparisons across languages and country contexts.
• Performance With over 519 million adjustable settings, this AI tool correctly identified ESG content with an accuracy of approximately 901 TP3T. It performed best in languages with more training data available, including English, Chinese, and Japanese.
  

3. Cross-language consistency analysis

• Of the 163 companies with bilingual reports, those from China, Japan, and Korea showed a high degree of consistency with the English version, suggesting effective translation and standardization of ESG messages.
• On the other hand, the Thai and Indonesian reports showed more pronounced discrepancies between the English and local language versions. It has been suggested that this may be due to translation practices and a lack of ESG literacy in the local languages, and that the scarcity of ESG-related content in these languages may have limited adequate training on the TMPT model.

Summary and Discussion

The analysis in this study provides valuable insights into ESG disclosure patterns among leading companies in East and Southeast Asia.

1. Overall ESG Emphasis Pattern

• In all reports reviewed, companies than environmental and social topics,Economic and Governance Topicsand especially more emphasis was placed on economic spillovers, production costs, and governance risks.
• Social topics such as domestic job creation and labor environment also received attention, while environmental topics such as mining, consumption, and greenhouse gases received moderate attention.
  

2. Major trends by country/region

• China: China The report has a strong focus on economic topics (e.g., spillover effects, production costs), which is consistent with China's focus on economic recovery after the 2023 novel coronavirus outbreak.
• South Korea: (English) The report focuses most heavily on environmental and social topics, such as the work environment and greenhouse gas emissions, but also noted significant variation within the market.
• Japan: Japan The report reflected a balanced focus, with the greatest emphasis on social aspects, particularly job creation and the working environment.
• Indonesia, Malaysia, Philippines:. In these economies, the reports generally focused more on community-related issues and human rights. Malaysia specifically emphasized governance risks, while the Philippines focused on water management.
• Singapore: Singapore The focus of the annual report was tilted toward economic topics, particularly job creation.
• Thailand: Thailand The report showed a relatively balanced profile, close to the regional average on most topics.
• Intra-market Variation:. While market-level trends are discernible, we also observed marked variations within each economy. For example, Korean firms vary widely in their focus on environmental and social issues.
  

3. Policy Implications

• Usefulness of AI:. This study shows that AI can be very useful in understanding subtle ESG reporting patterns across languages and economies. This provides a valuable alternative to traditional single scale assessments.
• Factors Affecting Disclosures: The influence of national priorities, regulatory environment, and cultural factors on ESG disclosure is also highlighted. For example, the emphasis on economic factors by Chinese companies reflects the government's drive toward economic stabilization, while Korean companies' involvement in social and environmental themes reflects social pressures and the influence of large corporations. Community-related disclosures in Southeast Asia may reflect local development challenges.
• Limitations and challenges of the AI model:. Limitations due to AI models being trained on linguistically uneven data have also been identified that may exacerbate gaps in ESG understanding and communication across economies. This suggests the need for capacity building in ESG education and data availability in less expressive languages.
• The Value and Challenges of Integrated Reporting:. Integrated reports are on the rise and can increase transparency and enhance stakeholder engagement. This study addresses the main limitations of integrated reports: lack of standardization and comparability, and increases the actual value of integrated reports with narratives that explain corporate strategy.
• Future recommendations:. Regulators and companies should adopt a variety of communication tools, including integrated reports, to further improve transparency and corporate governance and promote timely disclosure practices. For small and mid-sized companies, publishing integrated reports and translating them into English is often a challenge due to the cost, but regardless of the format of disclosure and communication, advanced technology can provide a way to assess a company's business strategy along with the various environmental and social interests of its stakeholders The following are some examples

The research methodology for this report was developed by aiESG, a service provided by aiESG, Inc.aiESG for IRWe can also provide this service at "aiESG, Inc. aiESG Inc. provides comprehensive support ranging from organizing ESG indicators to specific assistance for each company, so please feel free to contact us if you have any problems with ESG-related information disclosure or practices.

About aIESG's services:https://aiesg.co.jp/service/

RELATED:
aiESG President Managi co-authored a report published in the Asian Bond Monitor, published by the Asian Development Bank.