You upload a csv with multiple columns, and the unstructured input might have a keyword in one of the columns (like author, or model number) that isn't present in the primary search column but still helps identify the entity.
You can already use multiple columns to match - in passing the request you need to specify the columns in refine plus the columns in the CSV they would match to.
It would be awesome to see support for the type of weighted column data that microsoft uses for product matching
white paper here : http://research.microsoft.com/pubs/148339/offerMatching_kdd.pdf
You upload a csv with multiple columns, and the unstructured input might have a keyword in one of the columns (like author, or model number) that isn't present in the primary search column but still helps identify the entity.
Just an idea!