Refining the Semantic Data Model to Improve Accuracy
To effectively use vector search for semantic long-term memory in an intelligent application, you must optimize the semantic data model to the application’s needs. As the semantic data model uses vector embedding models and vector search, you must optimize the contents of the embedded data and the way the data is retrieved.
Refining the semantic data model can lead to significant improvements in retrieval accuracy and overall application performance. In retrieval-augmented generation (RAG) applications, an effective semantic data model serves as the foundation for a robust retrieval system, which directly informs the quality of the generated outputs. The rest of the chapter examines different ways in which you can refine the semantic data model and retrieval.
This chapter will cover the following topics:
- Experimenting with different embedding models
- Fine-tuning embedding models
- Including metadata in the...
