A knowledge base is an organized collection of information that can be searched to help answer questions. In conversational AI, it supplies relevant data when built-in model knowledge is insufficient or when answers should reflect specific documents. section 1.1.3
A search-based assistant can retrieve information from the knowledge base and use those results to generate an answer. This supports pure conversational search or a hybrid design in which common questions use intents while other questions go to search or generative AI. section 1.1.3
Treat the knowledge base as an operational source that requires ongoing review:
- Audit source documents and URLs to confirm that they contain the most relevant information. section 4.2.3
- Review supplemental document enrichments and the ingestion schedule so answers remain current. section 4.2.3
- Use feedback and test results to investigate whether weak understanding comes from the knowledge base or another conversational component. section 4.2.3
A hybrid question-answering assistant
An assistant can handle its most common questions through predefined intents, then search a knowledge base for less common questions and generate answers from the retrieved results. A stale source document or delayed ingestion schedule can make otherwise capable answers unreliable. section 1.1.3 section 4.2.3
