Closed dennlinger closed 2 years ago
Current implementation is inefficient because I think we're re-computing the reference sentences' n-grams for each sentence in the summary text.
If true, this could significantly speed up the computation, by an order of O(N).
Problem actually existed and could be solved. For future reference, would be better to offer a way with caching maybe? Closed with 6f17f50
Current implementation is inefficient because I think we're re-computing the reference sentences' n-grams for each sentence in the summary text.
If true, this could significantly speed up the computation, by an order of O(N).