Environmental, social, and corporate governance (ESG) refers to the three important contributors to the sustainable growth of firms. Firms publish corporate social responsibility (CSR) reports that include quantitative and qualitative information concerning ESG activities. Although these reports are easily accessed, their qualitative information is hard to apply because manual analyses are difficult. We develop a text mining model that visualizes ESG activities from the structure of ESG-related words in CSR reports. This model quickly, effectively, and objectively facilitates processing CSR reports and comparing them with reports from peer firms. We analyze Japanese CSR reports and present an example. Further, we propose scores to evaluate the quantity and specificity of ESG activities. From the result, we obtain the following findings. First, large quantity and high specificity of ESG activities indicate a higher current ESG quantitative performance. Second, the high specificity of E-related and the large quantity of S and G-related activities portend subsequent improvement of ESG quantitative performance.

中文翻译：

---

评估企业 ESG 活动的文本挖掘模型：日本企业的应用

环境、社会和公司治理（ESG）是指企业可持续发展的三个重要因素。公司发布企业社会责任 (CSR) 报告，其中包括有关 ESG 活动的定量和定性信息。虽然这些报告很容易访问，但它们的定性信息很难应用，因为手动分析很困难。我们开发了一个文本挖掘模型，该模型根据 CSR 报告中 ESG 相关词的结构将 ESG 活动可视化。该模型快速、有效和客观地促进了 CSR 报告的处理，并将其与同行公司的报告进行比较。我们分析了日本的 CSR 报告并提供了一个例子。此外，我们提出评分来评估 ESG 活动的数量和特异性。从结果中，我们得到以下发现。第一的，ESG 活动的大量和高特异性表明当前 ESG 定量表现更高。其次，E 相关的高度特异性和大量的 S 和 G 相关活动预示着 ESG 量化绩效的后续改进。