Added parameter to allow negative examples in ClassificationCorpus

flair/datasets/document_classification.py

@@ -37,6 +37,7 @@ def __init__(
             memory_mode: str = "partial",
             label_name_map: Dict[str, str] = None,
             skip_labels: List[str] = None,
+            allow_examples_without_labels=False,
             encoding: str = 'utf-8',
     ):
         """
@@ -55,6 +56,7 @@ def __init__(
         if full corpus and all embeddings fits into memory for speedups during training. Otherwise use 'partial' and if
         even this is too much for your memory, use 'disk'.
         :param label_name_map: Optionally map label names to different schema.
+        :param allow_examples_without_labels: set to True to allow Sentences without label in the corpus.
         :param encoding: Default is 'uft-8' but some datasets are in 'latin-1
         :return: a Corpus with annotated train, dev and test data
         """
@@ -73,6 +75,7 @@ def __init__(
             memory_mode=memory_mode,
             label_name_map=label_name_map,
             skip_labels=skip_labels,
+            allow_examples_without_labels=allow_examples_without_labels,
             encoding=encoding,
         )
 
@@ -87,6 +90,7 @@ def __init__(
             memory_mode=memory_mode,
             label_name_map=label_name_map,
             skip_labels=skip_labels,
+            allow_examples_without_labels=allow_examples_without_labels,
             encoding=encoding,
         ) if test_file is not None else None
 
@@ -101,6 +105,7 @@ def __init__(
             memory_mode=memory_mode,
             label_name_map=label_name_map,
             skip_labels=skip_labels,
+            allow_examples_without_labels=allow_examples_without_labels,
             encoding=encoding,
         ) if dev_file is not None else None
 
@@ -125,6 +130,7 @@ def __init__(
             memory_mode: str = "partial",
             label_name_map: Dict[str, str] = None,
             skip_labels: List[str] = None,
+            allow_examples_without_labels=False,
             encoding: str = 'utf-8',
     ):
         """
@@ -143,6 +149,7 @@ def __init__(
         if full corpus and all embeddings fits into memory for speedups during training. Otherwise use 'partial' and if
         even this is too much for your memory, use 'disk'.
         :param label_name_map: Optionally map label names to different schema.
+        :param allow_examples_without_labels: set to True to allow Sentences without label in the Dataset.
         :param encoding: Default is 'uft-8' but some datasets are in 'latin-1
         :return: list of sentences
         """
@@ -169,14 +176,15 @@ def __init__(
         self.truncate_to_max_tokens = truncate_to_max_tokens
         self.filter_if_longer_than = filter_if_longer_than
         self.label_name_map = label_name_map
+        self.allow_examples_without_labels = allow_examples_without_labels
 
         self.path_to_file = path_to_file
 
         with open(str(path_to_file), encoding=encoding) as f:
             line = f.readline()
             position = 0
             while line:
-                if "__label__" not in line or (" " not in line and "\t" not in line):
+                if ("__label__" not in line and not allow_examples_without_labels) or (" " not in line and "\t" not in line):
                     position = f.tell()
                     line = f.readline()
                     continue
@@ -219,7 +227,7 @@ def __init__(
                     text = line[l_len:].strip()
 
                     # if so, add to indices
-                    if text and label:
+                    if text and (label or allow_examples_without_labels):
 
                         if self.memory_mode == 'partial':
                             self.lines.append(line)
@@ -257,7 +265,7 @@ def _parse_line_to_sentence(
         if self.truncate_to_max_chars > 0:
             text = text[: self.truncate_to_max_chars]
 
-        if text and labels:
+        if text and (labels or self.allow_examples_without_label):
             sentence = Sentence(text, use_tokenizer=tokenizer)
 
             for label in labels:

tests/resources/tasks/multi_class_negative_examples/dev.txt

@@ -0,0 +1,5 @@
+__label__apple apple
+__label__tv tv
+__label__guitar guitar
+__label__apple __label__tv apple tv
+ dev example without labels

tests/resources/tasks/multi_class_negative_examples/test.txt

@@ -0,0 +1,6 @@
+__label__guitar guitar
+__label__apple apple
+__label__tv tv
+__label__apple __label__tv apple tv
+__label__apple __label__guitar apple tv
+test example without labels

tests/resources/tasks/multi_class_negative_examples/train.txt

@@ -0,0 +1,8 @@
+__label__tv tv
+__label__apple __label__tv apple tv
+__label__apple apple
+__label__tv tv
+__label__apple __label__tv apple tv
+__label__guitar guitar
+__label__guitar guitar
+train example without labels

tests/test_data.py

@@ -704,6 +704,22 @@ def test_tagged_corpus_downsample():
     assert 3 == len(corpus.train)
 
 
+def test_classification_corpus_multi_labels_without_negative_examples(tasks_base_path):
+    corpus = flair.datasets.ClassificationCorpus(tasks_base_path / "multi_class_negative_examples",
+                                                 allow_examples_without_labels=False)
+    assert len(corpus.train) == 7
+    assert len(corpus.dev) == 4
+    assert len(corpus.test) == 5
+
+
+def test_classification_corpus_multi_labels_with_negative_examples(tasks_base_path):
+    corpus = flair.datasets.ClassificationCorpus(tasks_base_path / "multi_class_negative_examples",
+                                                 allow_examples_without_labels=True)
+    assert len(corpus.train) == 8
+    assert len(corpus.dev) == 5
+    assert len(corpus.test) == 6
+
+
 def test_spans():
     sentence = Sentence("Zalando Research is located in Berlin .")