change of behaviour for empty

README.md

@@ -62,6 +62,7 @@ It uses the same definition as in previous case, but it works with multiset of l
 - **predictions** *(list[Union[int,str]]): list of predictions to score. Each predictions should be a list of predicted labels*
 - **references** *(list[Union[int,str]]): list of reference for each prediction. Each reference should be a list of reference labels*
 
+
 ### Output Values
 
 This metric outputs a dictionary, containing:
@@ -70,6 +71,17 @@ This metric outputs a dictionary, containing:
 - accuracy
 - fscore
 
+
+Ff prediction and reference are empty lists, the output will be:
+```python
+{
+    "precision": 1.0,
+    "recall": 1.0,
+    "accuracy": 1.0,
+    "fscore": 1.0
+}
+```
+
 ## Citation
 
 ```bibtex

multi_label_precision_recall_accuracy_fscore.py

@@ -104,6 +104,9 @@ class MultiLabelPrecisionRecallAccuracyFscore(evaluate.Metric):
         )
 
     def eval_example(self, prediction, reference):
+        if len(prediction) == 0 and len(reference) == 0:
+            return 1, 1, 1
+
         if self.use_multiset:
             prediction = Counter(prediction)
             reference = Counter(reference)

tests.py

@@ -58,10 +58,10 @@ class MultiLabelPrecisionRecallAccuracyFscoreTest(TestCase):
     def test_empty(self):
         self.assertDictEqual(
             {
-                "precision": 0.0,
-                "recall": 0.0,
-                "accuracy": 0.0,
-                "fscore": 0.0
+                "precision": 1.0,
+                "recall": 1.0,
+                "accuracy": 1.0,
+                "fscore": 1.0
             },
             self.multi_label_precision_recall_accuracy_fscore.compute(
                 predictions=[