Add element confidence to CreateElementTranscriptions

arkindex/documents/api/ml.py

@@ -253,6 +253,7 @@ class ElementTranscriptionsBulk(CreateAPIView):
                     mirrored=self.element.mirrored,
                     worker_version=worker_version,
                     worker_run=worker_run,
+                    confidence=annotation.get('element_confidence'),
                 )
                 # Specify the annotated element has been created
                 annotation['created'] = True

arkindex/documents/serializers/ml.py

@@ -381,6 +381,11 @@ class SimpleTranscriptionSerializer(serializers.Serializer):
         required=False,
     )
     orientation = EnumField(TextOrientation, default=TextOrientation.HorizontalLeftToRight, required=False)
+    element_confidence = serializers.FloatField(
+        min_value=0,
+        max_value=1,
+        required=False,
+    )
 
     def validate(self, data):
         if 'confidence' not in data:

arkindex/documents/tests/test_bulk_element_transcriptions.py

@@ -148,6 +148,66 @@ class TestBulkElementTranscriptions(FixtureAPITestCase):
             ]
         )
 
+    def test_bulk_transcriptions_element_confidence(self):
+        """
+        Bulk creates a list of elements, with a confidence score and an attached transcription
+        """
+        # Create a manual transcription on the element
+        self.line.transcriptions.create(text='A manual transcription')
+        self.assertEqual(self.line.transcriptions.count(), 1)
+
+        existing_element_ids = list(Element.objects.get_descending(self.page.id).values_list('id', flat=True))
+
+        transcriptions = [
+            ([[13, 37], [133, 37], [133, 137], [13, 137], [13, 37]], 'Hello world !', 0.1337, 'vertical-lr', 0.5),
+            ([[24, 42], [64, 42], [64, 142], [24, 142], [24, 42]], 'I <3 JavaScript', 0.42, 'horizontal-lr', 0.7),
+        ]
+        data = {
+            'element_type': 'text_line',
+            'worker_run_id': str(self.worker_run.id),
+            'transcriptions': [{
+                'polygon': poly,
+                'text': text,
+                'confidence': confidence,
+                'orientation': orientation,
+                'element_confidence': element_confidence
+            } for poly, text, confidence, orientation, element_confidence in transcriptions]
+        }
+        self.client.force_login(self.user)
+        response = self.client.post(
+            reverse('api:element-transcriptions-bulk', kwargs={'pk': self.page.id}),
+            format='json',
+            data=data
+        )
+        self.assertEqual(response.status_code, status.HTTP_201_CREATED)
+
+        created_elts = Element.objects.get_descending(self.page.id).exclude(id__in=existing_element_ids)
+        self.assertEqual(created_elts.count(), 2)
+        self.assertTrue(all(map(lambda elt: elt.image == self.page.image, created_elts)))
+        self.assertListEqual(
+            [
+                (elt.paths.first().ordering, elt.name, elt.polygon.coords, elt.worker_version_id, elt.worker_run_id, elt.confidence)
+                for elt in created_elts
+            ],
+            [
+                (1, '2', ((13, 37), (13, 137), (133, 137), (133, 37), (13, 37)), self.worker_version.id, self.worker_run.id, 0.5),
+                (2, '3', ((24, 42), (24, 142), (64, 142), (64, 42), (24, 42)), self.worker_version.id, self.worker_run.id, 0.7)
+            ]
+        )
+        self.assertCountEqual(
+            created_elts.values_list(
+                'transcriptions__text',
+                'transcriptions__worker_version',
+                'transcriptions__worker_run',
+                'transcriptions__orientation',
+                'transcriptions__confidence'
+            ),
+            [
+                ('Hello world !', self.worker_version.id, self.worker_run.id, TextOrientation.VerticalLeftToRight, 0.1337),
+                ('I <3 JavaScript', self.worker_version.id, self.worker_run.id, TextOrientation.HorizontalLeftToRight, 0.42)
+            ]
+        )
+
     @override_settings(ARKINDEX_FEATURES={'search': False})
     def test_bulk_transcriptions_load(self):
         """