Rename QuestionAnswering template to QuestionAnsweringExtractive

src/datasets/tasks/__init__.py

@@ -2,16 +2,19 @@
 
 from ..utils.logging import get_logger
 from .base import TaskTemplate
-from .question_answering import QuestionAnswering
+from .question_answering import QuestionAnsweringExtractive
 from .text_classification import TextClassification
 
 
-__all__ = ["TaskTemplate", "QuestionAnswering", "TextClassification"]
+__all__ = ["TaskTemplate", "QuestionAnsweringExtractive", "TextClassification"]
 
 logger = get_logger(__name__)
 
 
-NAME2TEMPLATE = {QuestionAnswering.task: QuestionAnswering, TextClassification.task: TextClassification}
+NAME2TEMPLATE = {
+    QuestionAnsweringExtractive.task: QuestionAnsweringExtractive,
+    TextClassification.task: TextClassification,
+}
 
 
 def task_template_from_dict(task_template_dict: dict) -> Optional[TaskTemplate]:

src/datasets/tasks/question_answering.py

@@ -6,9 +6,9 @@
 
 
 @dataclass(frozen=True)
-class QuestionAnswering(TaskTemplate):
+class QuestionAnsweringExtractive(TaskTemplate):
     # `task` is not a ClassVar since we want it to be part of the `asdict` output for JSON serialization
-    task: str = "question-answering"
+    task: str = "question-answering-extractive"
     input_schema: ClassVar[Features] = Features({"question": Value("string"), "context": Value("string")})
     label_schema: ClassVar[Features] = Features(
         {

tests/test_arrow_dataset.py

@@ -20,7 +20,7 @@
 from datasets.info import DatasetInfo
 from datasets.splits import NamedSplit
 from datasets.table import ConcatenationTable, InMemoryTable, MemoryMappedTable
-from datasets.tasks import QuestionAnswering, TextClassification
+from datasets.tasks import QuestionAnsweringExtractive, TextClassification
 from datasets.utils.logging import WARNING
 
 from .conftest import s3_test_bucket_name
@@ -739,7 +739,9 @@ def test_concatenate_with_equal_task_templates(self, in_memory):
 
     def test_concatenate_with_mixed_task_templates_in_common(self, in_memory):
         tc_template = TextClassification(text_column="text", label_column="labels")
-        qa_template = QuestionAnswering(question_column="question", context_column="context", answers_column="answers")
+        qa_template = QuestionAnsweringExtractive(
+            question_column="question", context_column="context", answers_column="answers"
+        )
         info1 = DatasetInfo(
             task_templates=[qa_template],
             features=Features(
@@ -794,7 +796,9 @@ def test_concatenate_with_mixed_task_templates_in_common(self, in_memory):
     def test_concatenate_with_no_mixed_task_templates_in_common(self, in_memory):
         tc_template1 = TextClassification(text_column="text", label_column="labels")
         tc_template2 = TextClassification(text_column="text", label_column="sentiment")
-        qa_template = QuestionAnswering(question_column="question", context_column="context", answers_column="answers")
+        qa_template = QuestionAnsweringExtractive(
+            question_column="question", context_column="context", answers_column="answers"
+        )
         info1 = DatasetInfo(
             features=Features(
                 {
@@ -2107,7 +2111,7 @@ def test_task_question_answering(self, in_memory):
                 ),
             }
         )
-        task = QuestionAnswering(
+        task = QuestionAnsweringExtractive(
             context_column="input_context", question_column="input_question", answers_column="input_answers"
         )
         info = DatasetInfo(features=features_before_cast, task_templates=task)
@@ -2124,13 +2128,13 @@ def test_task_question_answering(self, in_memory):
                     set(dset.flatten().column_names),
                 )
                 self.assertDictEqual(features_before_cast, dset.features)
-                with dset.prepare_for_task(task="question-answering") as dset:
+                with dset.prepare_for_task(task="question-answering-extractive") as dset:
                     self.assertSetEqual(
                         set(["context", "question", "answers.text", "answers.answer_start"]),
                         set(dset.flatten().column_names),
                     )
                     self.assertDictEqual(features_after_cast, dset.features)
-        # Test we can load from QuestionAnswering template
+        # Test we can load from QuestionAnsweringExtractive template
         info.task_templates = None
         with tempfile.TemporaryDirectory() as tmp_dir, Dataset.from_dict(data, info=info) as dset:
             with dset.prepare_for_task(task=task) as dset:

tests/test_tasks.py

@@ -1,7 +1,7 @@
 from unittest.case import TestCase
 
 from datasets.features import ClassLabel, Features, Sequence, Value
-from datasets.tasks import QuestionAnswering, TextClassification
+from datasets.tasks import QuestionAnsweringExtractive, TextClassification
 
 
 class TextClassificationTest(TestCase):
@@ -25,7 +25,7 @@ def test_from_dict(self):
 
 class QuestionAnsweringTest(TestCase):
     def test_column_mapping(self):
-        task = QuestionAnswering(
+        task = QuestionAnsweringExtractive(
             context_column="input_context", question_column="input_question", answers_column="input_answers"
         )
         self.assertDictEqual(
@@ -49,7 +49,7 @@ def test_from_dict(self):
             "question_column": "input_question",
             "answers_column": "input_answers",
         }
-        task = QuestionAnswering.from_dict(template_dict)
-        self.assertEqual("question-answering", task.task)
+        task = QuestionAnsweringExtractive.from_dict(template_dict)
+        self.assertEqual("question-answering-extractive", task.task)
         self.assertEqual(input_schema, task.input_schema)
         self.assertEqual(label_schema, task.label_schema)