BLEU score refactoring.

math.fsum is used to make float sum stable.

nltk/align/bleu.py

@@ -31,7 +31,7 @@ class BLEU(object):
 
     Part 2 - brevity penalty
 
-    As the modified n-gram precision stil has the problem from the short
+    As the modified n-gram precision still has the problem from the short
     length sentence, brevity penalty is used to modify the overall BLEU
     score according to length.
 
@@ -89,34 +89,24 @@ class BLEU(object):
 
     @staticmethod
     def compute(candidate, references, weights):
+        candidate = [c.lower() for c in candidate]
+        references = [[r.lower() for r in reference] for reference in references]
 
-        candidate = list(map(lambda x: x.lower(), candidate))
-        references = list(map(lambda x: [c.lower() for c in x], references))
-
-        n = len(weights)
+        p_ns = (BLEU.modified_precision(candidate, references, i) for i, _ in enumerate(weights, start=1))
+        s = math.fsum(w * math.log(p_n) for w, p_n in zip(weights, p_ns))
 
         bp = BLEU.brevity_penalty(candidate, references)
-
-        s = 0.0
-        i = 1
-        for weight in weights:			
-            p_n = BLEU.modified_precision(candidate, references, i)
-            if p_n != 0:
-                s += weight * math.log(p_n)
-            i += 1
-
         return bp * math.exp(s)
 
     @staticmethod
     def modified_precision(candidate, references, n):
 
         candidate_ngrams = list(ngrams(candidate, n))
+        c_words = set(candidate_ngrams)
 
-        if len(candidate_ngrams) == 0:
+        if not c_words:
             return 0
 
-        c_words = set(candidate_ngrams)
-
         for word in c_words:
             count_w = candidate_ngrams.count(word) + 1
 
@@ -125,23 +115,22 @@ def modified_precision(candidate, references, n):
                 reference_ngrams = list(ngrams(reference, n))
 
                 count = reference_ngrams.count(word) + 1
-                if count > count_max:
-                    count_max = count
+                count_max = max(count, count_max)
 
+        # TODO: count_w == candidate_ngrams.count(c_words[-1]) + 1
+        # (even though c_words is a set, so there is no last element, it's the last element returned by the iterator.)
+        # Is it the desired behavior?
         return min(count_w, count_max) / (len(candidate) + len(c_words))
 
     @staticmethod
     def brevity_penalty(candidate, references):
         c = len(candidate)
-
-        lengthes_ref = map(lambda x: abs(len(x) - c), references)
-
-        r = min(*lengthes_ref)
+        r = min(abs(len(r) - c) for r in references)
 
         if c > r:
             return 1
         else:
-            return math.exp(1 - r/c)
+            return math.exp(1 - r / c)
 
 # run doctests
 if __name__ == "__main__":