Enhancing Business Analytics through Generative AI: Integrating Large Language Models with Proprietary Knowledge Graphs for Advanced Data Querying and Visualization

While the generally available large language models (LLMs), pre-trained on large, general-purpose datasets, are easy to use, they require additional fine-tuning to perform in an academic or business context requiring specific domain knowledge. The current chapter demonstrates putting a transformer-based LLM in context. The example shows a solution for getting insights from a vast internal knowledge base, a unique source of graph-structured domain-specific information. Typically, querying a knowledge graph (a graph database) requires specific skills, limiting access to its data. The use of generative AI enables querying the graph using natural language. An LLM, fine-tuned with the knowledge graph ontology, translates natural language questions to programmatic queries to the graph database and shows the response in human-readable, graphically enriched form. The authors present an approach they have tested and proved successful in one of the leading media companies in the world. With the help of a concrete showcase from the finance and business news domain, they describe a methodology that is universally applicable to various industries and educational environments. Also, the established solution is vendor-independent and can be built on different underlying technologies, services, and platforms. The achieved outcome unveils the full potential of the data enabling advanced analytics and can be instrumental in higher education for studies in applied business analytics.