Question

A financial research firm uses an Amazon Bedrock Knowledge Base to answer analyst questions about market reports. Reports are published continuously, but analysts frequently complain that the assistant cites outdated figures from reports published over a year ago when newer versions exist. The team wants queries to prioritize recent documents without deleting historical data, which must remain searchable for compliance reviews. What is the most effective way to address this?

Accepted Answer

Attach a 'publish_date' metadata attribute to each document during ingestion and apply metadata filtering at query time to restrict or bias retrieval toward recent dates.

Answer

Reduce the chunk size in the Knowledge Base so that newer, smaller chunks rank higher in vector similarity scoring.

Answer

Delete all documents older than one year from the S3 data source and re-sync the Knowledge Base.

Answer

Switch the embedding model to a higher-dimension model to improve the ranking of recent documents.