Formatting

metrics/bleu.py

@@ -47,8 +47,7 @@ def _get_ngrams(segment, max_order):
     return ngram_counts
 
 
-def compute_bleu(reference_corpus, translation_corpus, max_order=4,
-                 smooth=True):
+def compute_bleu(reference_corpus, translation_corpus, max_order=4, smooth=True):
     """Computes BLEU score of translated segments against one or more references.
 
     Args:
@@ -67,8 +66,7 @@ def compute_bleu(reference_corpus, translation_corpus, max_order=4,
     possible_matches_by_order = [0] * max_order
     reference_length = 0
     translation_length = 0
-  for (references, translation) in zip(reference_corpus,
-                                       translation_corpus):
+    for (references, translation) in zip(reference_corpus, translation_corpus):
         reference_length += min(len(r) for r in references)
         translation_length += len(translation)
 
@@ -87,17 +85,19 @@ def compute_bleu(reference_corpus, translation_corpus, max_order=4,
     precisions = [0] * max_order
     for i in range(0, max_order):
         if smooth:
-      precisions[i] = ((matches_by_order[i] + 1.) /
-                       (possible_matches_by_order[i] + 1.))
+            precisions[i] = (matches_by_order[i] + 1.0) / (
+                possible_matches_by_order[i] + 1.0
+            )
         else:
             if possible_matches_by_order[i] > 0:
-        precisions[i] = (float(matches_by_order[i]) /
-                         possible_matches_by_order[i])
+                precisions[i] = (
+                    float(matches_by_order[i]) / possible_matches_by_order[i]
+                )
             else:
                 precisions[i] = 0.0
 
     if min(precisions) > 0:
-    p_log_sum = sum((1. / max_order) * math.log(p) for p in precisions)
+        p_log_sum = sum((1.0 / max_order) * math.log(p) for p in precisions)
         geo_mean = math.exp(p_log_sum)
     else:
         geo_mean = 0
@@ -105,14 +105,21 @@ def compute_bleu(reference_corpus, translation_corpus, max_order=4,
     ratio = float(translation_length) / reference_length
 
     if ratio > 1.0:
-    bp = 1.
+        bp = 1.0
     else:
-    bp = math.exp(1 - 1. / ratio)
+        bp = math.exp(1 - 1.0 / ratio)
     bleu = geo_mean * bp
-  print(geo_mean)
-  bleu_score_dict = {"bleu":bleu,"precision":precisions,"bp":bp,"ratio":ratio,"trans_len":translation_length,"ref_len":reference_length}
+    bleu_score_dict = {
+        "bleu": bleu,
+        "precision": precisions,
+        "bp": bp,
+        "ratio": ratio,
+        "trans_len": translation_length,
+        "ref_len": reference_length,
+    }
     return bleu_score_dict  # (bleu, precisions, bp, ratio, translation_length, reference_length)
 
+
 def bleu_test_case():
     """A simple functionality test case to evaluate BLEU"""
     generated = [[["a", "=", "b", "\n", "y", "=", "a", "+", "1"]]]
@@ -120,6 +127,7 @@ def bleu_test_case():
     score_dict = compute_bleu(generated, reference, smooth=False)
     return score_dict
 
+
 if __name__ == "__main__":
     score_dict = bleu_test_case()
     print(score_dict)

metrics/extrinsic_eval.py

@@ -12,7 +12,10 @@ def compute_exact_match(references,generated)->float:
     returns:
         exact_match_accuracy : Float
     """
-    assert(len(references[0])==len(generated),"Number of Samples should be equal in References and Synthesized Outputs..")
+    assert (
+        len(references[0]) == len(generated),
+        "Number of Samples should be equal in References and Synthesized Outputs..",
+    )
     exact_match_count = 0.0
     for gen, ref in zip(generated, references[0]):
         if gen == ref:
@@ -20,6 +23,7 @@ def compute_exact_match(references,generated)->float:
     exact_match_acc = exact_match_count / len(generated)
     return exact_match_acc
 
+
 def compute_metrics(references, generated) -> dict:
     """
     Calculates various metrics and returns the calculated dict of these matrics.
@@ -31,7 +35,11 @@ def compute_metrics(references,generated) -> dict:
     returns:
         A dicitonary with different metrics intact.
     """
-    metrics_dict = {} #Update as in new metrics are added over here.
+    metrics_dict = {
+        "smoothed_bleu_4": None,
+        "blue_4": None,
+        "exact_match_acc": None,
+    }  # Update as in new metrics are computed.
     metrics_dict["smoothed_bleu_4"] = compute_bleu(references, generated, smooth=True)
     metrics_dict["bleu_4"] = compute_bleu(references, generated, smooth=False)
     metrics_dict["exact_match_acc"] = compute_exact_match(references, generated)