I want to perform some stats on terms in a text field. To do that, I think the idea is to use a term aggregation on the wanted field. If I want top terms, I need to have this string analyzed. However, as my field is very big, I hit the field data cache circuit breaker when running this aggregation. I know that ES can now use doc values to limit RAM impact for aggregation computing but it seems that we cannot use it with analyzed field. Does anybody have any idea on how to achieve what I want to do?