Add weight sum result

action_generation.py

@@ -47,8 +47,8 @@ Examples:
     Examples should be written in doctest format, and should illustrate how
     to use the function.
 
-    >>> my_new_module = evaluate.load("my_new_module")
-    >>> results = my_new_module.compute(references=[0, 1], predictions=[0, 1])
+    >>> metric = evaluate.load("DarrenChensformer/aciton_generation")
+    >>> results = metric.compute(references=[0, 1], predictions=[0, 1])
     >>> print(results)
     {'accuracy': 1.0}
 """
@@ -56,6 +56,101 @@ Examples:
 # TODO: Define external resources urls if needed
 BAD_WORDS_URL = "http://url/to/external/resource/bad_words.txt"
 
+class BaseEvaluater:
+    eps = 1e-8
+
+    def __call__(self, preds, labels):
+        return self._compute(preds, labels)
+
+    def _compute(self, preds, labels):
+        # calculate precision, recall, f1
+        tp, fp, fn = 0, 0, 0
+        for pred, label in zip(preds, labels):
+            tp += len(set(pred) & set(label))
+            fp += len(set(pred) - set(label))
+            fn += len(set(label) - set(pred))
+        precision = tp / (tp + fp + self.eps)
+        recall = tp / (tp + fn + self.eps)
+        f1 = 2 * precision * recall / (precision + recall)
+
+        return {
+            "precision": round(precision, 4),
+            "recall": round(recall, 4),
+            "f1": round(f1, 4)
+        }
+
+class ClassEvaluater(BaseEvaluater):
+    def __init__(self, valid_labels=None):
+        self.valid_labels = valid_labels
+
+    def __call__(self, preds, labels):
+        preds = map(self.extract_class, preds)
+        labels = map(self.extract_class, labels)
+        # helper function to extract valid tags
+        preds = list(map(self.extract_valid, preds))
+        labels = list(map(self.extract_valid, labels))
+        return self._compute(preds, labels)
+
+    def extract_valid(self, tags):
+        # TODO: if valid_labels is None:
+        tags = list(filter(lambda tag: tag in self.valid_labels, tags))
+        return tags
+
+    def extract_class(self, tags):
+        tags = map(lambda tag: tag.replace("/ ", "/"), tags)
+        tags = list(map(self.batch_extract_class, tags))
+        # deduplicate
+        tags = list(dict.fromkeys(tags))
+        return tags
+
+    def batch_extract_class(self, tag):
+        # filter out invalid tags
+        tag = tag.split('/')
+        if len(tag)==3:
+            _class = '/'.join(tag[:2])
+        elif len(tag)==4:
+            _class = '/'.join(tag[:3])
+        elif len(tag)==1:
+            _class = ''
+        else:
+            _class = None
+        if _class in self.valid_labels:
+            return _class
+        else:
+            return ""
+
+
+class PhraseEvaluater(BaseEvaluater):
+    def __init__(self, valid_labels=None):
+        self.valid_labels = valid_labels
+
+    def __call__(self, preds, labels):
+        preds = map(self.extract_phrase, preds)
+        labels = map(self.extract_phrase, labels)
+        return self._compute(preds, labels)
+    
+    def extract_phrase(self, tags):
+        tags = map(lambda tag: tag.replace("/ ", "/"), tags)
+        tags = list(map(self.batch_extract_phrase, tags))
+        # deduplicate
+        tags = list(dict.fromkeys(tags))
+        return tags
+
+    def batch_extract_phrase(self, phrase):
+        # filter out invalid tags
+        tag = phrase.split('/')
+        if len(tag)==3:
+            _class = '/'.join(tag[:2])
+        elif len(tag)==4:
+            _class = '/'.join(tag[:3])
+        elif len(tag)==1:
+            _class = ''
+        else:
+            _class = None
+        if _class in self.valid_labels:
+            return phrase.replace(_class, '')
+        else:
+            return ""
 
 @evaluate.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
 class action_generation(evaluate.Metric):
@@ -71,8 +166,8 @@ class action_generation(evaluate.Metric):
             inputs_description=_KWARGS_DESCRIPTION,
             # This defines the format of each prediction and reference
             features=datasets.Features({
-                'predictions': datasets.Value('int64'),
-                'references': datasets.Value('int64'),
+                'predictions': datasets.Sequence(datasets.Value('string')),
+                'references': datasets.Sequence(datasets.Value('string')),
             }),
             # Homepage of the module for documentation
             homepage="http://module.homepage",
@@ -86,10 +181,25 @@ class action_generation(evaluate.Metric):
         # TODO: Download external resources if needed
         pass
 
-    def _compute(self, predictions, references):
+    def _compute(self, predictions, references,
+            valid_labels=None, detailed_scores=False,
+            weights={"class": 0.8, "phrase": 0.2}
+        ):
         """Returns the scores"""
-        # TODO: Compute the different scores of the module
-        accuracy = sum(i == j for i, j in zip(predictions, references)) / len(predictions)
-        return {
-            "accuracy": accuracy,
-        }
+        weights = {"class": 0.8, "phrase": 0.2}
+        class_eval = ClassEvaluater(valid_labels)(predictions, references)
+        phrase_eval = PhraseEvaluater(valid_labels)(predictions, references)
+        weight_sum = {
+            key: round((class_eval[key] * weights["class"]) + (phrase_eval[key] * weights["phrase"]), 4)
+            for key in class_eval
+        }
+        if detailed_scores:
+            results = {
+                "class": class_eval,
+                "phrase": phrase_eval,
+                "weighted_sum": weight_sum
+            }
+        else:
+            results = weight_sum
+
+        return results