This presentation will report on the integration of LLMs, semantic knowledge models and recommender systems that serve as the enabling technology for a Search Generative Experience (SGE). The presented application is at its core based on a comprehensive ESG domain model to provide sustainable investment and ESG practitioners with referenceable and reliable information.
Firstly, we address the question of what a practical search or question and answer engine should look like from the user’s perspective, especially when dealing with a complex domain such as ESG. How can users with different levels of knowledge and different roles and tasks be served and learn to understand relationships across a domain without losing the common thread? How can it be ensured that authoritative sources are always referenced, if they are available? How can this design pattern be applied to other application classes, e.g. when setting up a help and support desk for products that require explanation? In the second part of the presentation, we will discuss the underlying architecture and show how knowledge models provide key services for prompt engineering and retrieval-augmented generation (RAG). In addition, we take a closer look at the underlying ESG knowledge model and learn about the principles according to which it is developed and how LLMs are used for this purpose. We briefly return to the roots of the Semantic Web and look at the data model from the perspectives of interoperability, explainability and the possibility of developing semantic models independently of the application logic and thus making them reusable.
Finally, we take a bird’s eye view and look at the presented SGE for ESG application as one facet of a more comprehensive knowledge hub. We see the underlying knowledge graph as the driver for the efficient development of domain-specific applications, e.g. to enable specific views of ESG data from investors or enterprises in the case of sustainable business.
Date:
May 8, 11:00 AM-11:30 AM
Classroom 215
Speaker:
Andreas Blumauer, Semantic Web Company
