Generate Text based on Proprietary, Specific Knowledge

Food for Thought

Applications can extend the context provided to AI foundation models with information coming from proprietary, specific databases, enterprise applications or unstructured documents to generate text for different purposes (writing articles or reports, suggesting ideas or questions for a defined topic, etc.) based both on public knowledge that was available until the model was trained and specific, proprietary information coming from proprietary databases, applications, and documents.

Example

A BTP application compiles a sustainability report combining information stored in proprietary databases, applications, and documents.

Key Questions

• What text needs to be generated looking at any specific, company internal information?
• Are those text based on real-time, updated information?

Implementation

To implement this AI capability, use a Retrieval Augmented Generation approach. Below the related Best Practices.

Related Best Practices

• Vector RAG vs Graph RAG
• Vector-based RAG (1/2) Embedding
• Vector-based RAG (2/2) Query Pipeline
• Knowledge Graph RAG (1/2) Creation
• Knowledge Graph RAG (2/2) Query Pipeline