Document level metrics

[yemaedahrav]

I notice that there are two types of metrics, one is token-level and other is retrieval level, where check for the golden answer snippets is looked for in retrieved documents. Are there plans for adding document level retrievals? We measure metrics on the retrieved documents against the ground truth source documents. If the dataset is such. This is useful in a lot of scenarios. I can implement this maybe.

[ignorejjj]

This is a great idea, but due to the limitations of our own dataset (lack of golden documents), we do not have a similar plan. If you can implement one, we would be happy to add it to the library.