Avoid looping through the whole counter in bleu_score method (#1913)

* avoid to loop through the whole counter in bleu_score method

* fix bug when max_n > len(candidate)

* add comment to explain L88

torchtext/data/metrics.py

@@ -65,11 +65,12 @@ def bleu_score(candidate_corpus, references_corpus, max_n=4, weights=[0.25] * 4)
     refs_len = 0.0
 
     for (candidate, refs) in zip(candidate_corpus, references_corpus):
-        candidate_len += len(candidate)
+        current_candidate_len = len(candidate)
+        candidate_len += current_candidate_len
 
         # Get the length of the reference that's closest in length to the candidate
         refs_len_list = [float(len(ref)) for ref in refs]
-        refs_len += min(refs_len_list, key=lambda x: abs(len(candidate) - x))
+        refs_len += min(refs_len_list, key=lambda x: abs(current_candidate_len - x))
 
         reference_counters = _compute_ngram_counter(refs[0], max_n)
         for ref in refs[1:]:
@@ -79,11 +80,12 @@ def bleu_score(candidate_corpus, references_corpus, max_n=4, weights=[0.25] * 4)
 
         clipped_counter = candidate_counter & reference_counters
 
-        for ngram in clipped_counter:
-            clipped_counts[len(ngram) - 1] += clipped_counter[ngram]
+        for ngram, count in clipped_counter.items():
+            clipped_counts[len(ngram) - 1] += count
 
-        for ngram in candidate_counter:  # TODO: no need to loop through the whole counter
-            total_counts[len(ngram) - 1] += candidate_counter[ngram]
+        for i in range(max_n):
+            # The number of N-grams in a `candidate` of T tokens is `T - (N - 1)`
+            total_counts[i] += max(current_candidate_len - i, 0)
 
     if min(clipped_counts) == 0:
         return 0.0