How to best create a dataset for FAQs

Question

We are considering creating an index of FAQs.

The dataset is going to have one-to-one answers to questions, but we are still researching if this is the best way to do it.

My concern is that in this case, we would only get one hit answer per question.

Would it be better to not put it in a one-to-one format so that we can get relevant background and other similar answers to a single question?

Note that a vector search is used.

Answer 1

@ryota.matsuda , Apologies for the delayed response.

Based on your requirement, you may enable multiple options with citations that will give the user the option to choose what perhaps is a second or third choice to what they are looking for.

Also, I would suggest you to enable semantic ranking that will add an extra layer of ranking for semantic related content for better results.