You can do set overlap on the extracted terms and that gives similar relevancy to the chunks as done by their embedding vectors. It’s just a lower dimensionally

You can do set overlap on the extracted terms and that gives similar relevancy to the chunks as done by their embedding vectors. It’s just a lower dimensionally of the relationships between the chunk and the query.