continuous-eval
continuous-eval copied to clipboard
Metadata comparison for retrieval metrics
@kelvinchanwh I actually had this issue open a while ago. I will find some time to implement it. I plan to do the below. Let me know if you have any suggestions.
TODO: Create a new metric PrecisionRecallF1_Metadata that compares the metadata of retrieved_context vs. metadata of ground_truth_context. metadata should be a json containing information such as can be document id / page number / etc.