Knowledge-hub.eco

A tool that makes knowledge about environmental, social, and corporate governance aspects easily available.

Background

Environmental, social, and corporate governance (ESG) aspects play an increasingly important role and have a substantial impact on investment strategies and consumer decisions. As a consequence, comprehensive and trustworthy ESG knowledge is more relevant than ever. For many people, however, access to this important information is associated with obstacles, especially for those with insufficient expert knowledge.

Project Content

In a joint project, experts of the Semantic Web Company from Vienna and the Institute for Innovation Systems of the St. Pölten University of Applied Sciences have developed the first AI-assisted question-and-answer machine that makes reliable and fact-based information on the complex field of ESG accessible in an intuitive manner. It was first presented to the public within the framework of the Long Night of Research at the St. Pölten UAS on 24 May 2024.

A beta version of the service is available at https://knowledge-hub.eco/.

Objectives and Methods

In a groundbreaking development step, the project team combined large language models (LLMs) and knowledge graphs (KGs) to create a high-performance tool that can be used to better structure and process the complex field of sustainability and ESG management. The objective was to make information more easily usable and to place it on a fact-based foundation.

The core piece of our method is the so-called semantic RAG approach (Retrieval-Augmented Generation) with whose help the output of a large language model (LLM) can be improved. This is achieved through the use of a sound knowledge base which massively reduces the training effort of the language model at the same time. Consequently, the search results are more reliable, and a lot of energy is saved in the operation of the artificial intelligence as well. Another great advantage of the RAG approach is that it can be used to tailor LLMs to specific domains or organisational knowledge bases, so that the relevance, accuracy, and usefulness of the generated results can be increased in a cost-effective manner.

Results

The project team has developed a tool that makes ESG knowledge readily accessible for laypersons. To this end, we have developed a general reference model for ESG that currently consists of 7,500 topic-relevant concepts and more than 40,000 terms. The resulting ESG graph portrays the dependencies between ESG stakeholders, regulations, clean technologies, and risks and opportunities that arise from ESG topics and forms the fundamental knowledge base for the AI machine which makes the information available on request.