Problem:
Talk to your textual data in a most effective way. One problem with large language models (LLM) like ChatGPT and chatbots based on them is that they can hallucinate and return great-sounding but wildly inaccurate results. The problem is that these LLMs are inherently black boxes, so it is hard to fix and retrain models to reduce hallucinations. Consequently, it might not be a good idea to depend on answers from the model if mission-critical tasks or lives are at stake. However, there is tremendous value in having the ability to interact with such chatbots and use them as an interface for various applications.

Solution:

• Making a structured representation of your textual data using knowledge graph. Use a LLM (eg. gpt3.5-turbo ) to generate a knowledge graph.
• Upload this data to a graph database (eg. Neo4j )
• Make an interface to this data using a pre-trained LLM: the model should "know: how to convert human request to a database query. For this purpose, we have to provide a number of pairs "(prompts, related database query)" to be used for the LLM training.
• Figure 1 shows a knowledge graph for some articles on Quantum Computing.