Accessing the Capabilities of KGs and LLMs in Mapping Indicators within Sustainability Reporting Standards

abstract

Sustainability reporting is gaining importance in response to climate change and the pursuit of social sustainable development. In preparing these reports, sustainability managers are tasked with identifying sustainability indicators across multiple reporting standards. However, the challenge arises to locate the corresponding indicators in another standard through keyword searches due to the inconsistency of the naming conventions and classifications across different standards. Knowledge graphs(KGs) offer a promising solution for mapping the concepts of sustainability reporting from diverse standards. Nonetheless, traditional approaches to construct KGs are often time and resource intensive. In this context, the advanced natural language understanding capabilities of the Large Language Models (LLMs) could be explored to comprehend the reporting standards. Additionally, the rich knowledge structure of KGs could be leveraged to enhance the retrieval of relevant document snippets that describe the indicators within these standards. Accordingly, we propose a framework aimed at accessing the capabilities of KGs and LLMs in mapping indicators within sustainability reporting standards. This paper presents our framework, details two exploratory experiments, and discuses the preliminary results.