A corpus contains exact product terms and semantically related descriptions. Which search design should be considered?

1. Random row selection
2. A database backup as the retrieval engine
3. Only a row identity constraint
4. A hybrid of appropriate full-text and vector signals