Hi everyone,
I'm new to Neo4j and currently working on a dynamic Neo4j database where parent-child relationships are created programmatically without a predefined model. Each node in the database has unique properties, and child nodes are linked to their respective parents. I’d like to implement a question-answering system using NLP to query this knowledge base, ideally in a way that resembles the RAG approach.
I’m using Python to interact with Neo4j and have built a basic QA system using LangChain. However, I'm running into issues with retrieving data. For example, if I query, "Give me the details on wafer 9," it doesn’t match correctly because Neo4j might store it as "Wafer9" (with capitalization differences or slight variations). Additionally, I’m struggling to handle more complex queries effectively. I’ve experimented with few-shot prompting, but due to the dynamic nature of the data, it only works for closely matched cases.
I would appreciate any advice on how to effectively use NLP to query and retrieve information based on the properties and relationships in this kind of dynamic database.
Thank you!