Analyzing and Visualizing Text Information in Corporate Sustainability Reports Using Natural Language Processing Methods

• Published in: Applied sciences 2022-06, Vol.12 (11), p.5614
• Main Authors: Kang, Hyewon, Kim, Jinho
• Format: Article
• Language: English
• Subjects: Business indicators; Data mining; Environmental impact; False information; Language; Natural language processing; Sentiment analysis; Software; Statistical analysis; Sustainability; Sustainability reporting; sustainability reports; Sustainable development; sustainable development goals; thematic analysis; Trends
• ISSN: 2076-3417; 2076-3417
• DOI: 10.3390/app12115614

Sustainability is a major contemporary issue that affects everyone. Many companies now produce an annual sustainability report, mainly intended for their stakeholders and the public, enumerating their goals and degrees of achievement regarding sustainable development. Although sustainability reports are an important resource to understand a company’s sustainability strategies and practices, the difficulty of extracting key information from dozens or hundreds of pages with sustainability and business jargon has highlighted the need for metrics to effectively measure the content of such reports. Accordingly, many researchers have attempted to analyze the concepts and messages from sustainability reports using various natural language processing (NLP) methods. In this study, we propose a novel approach that overcomes the shortcomings of previous studies. Using the sentence similarity method and sentiment analysis, the study clearly shows thematic practices and trends, as well as a significant difference in the balance of positive and negative information in the reports across companies. The results of sentiment analysis prove that the new approach of this study is very useful. It confirms that companies actively use the sustainability report to improve their positive image when they experience a crisis. It confirms that companies actively use the sustainability report to improve their positive image when they experience a crisis. The inferences gained from this method will not only help companies produce better reports that can be utilized effectively, but also provide researchers with ideas for further research. In the concluding section, we summarize the implications of our approach and discuss limitations and future research areas.