feat: set value_tag to canonical label value during prediction import

robotoff/insights/importer.py

@@ -24,7 +24,12 @@
     get_product_store,
     is_valid_image,
 )
-from robotoff.taxonomy import Taxonomy, get_taxonomy
+from robotoff.taxonomy import (
+    Taxonomy,
+    TaxonomyType,
+    get_taxonomy,
+    match_taxonomized_value,
+)
 from robotoff.utils import get_logger, text_file_iter
 from robotoff.utils.cache import CachedStore
 
@@ -566,6 +571,17 @@ def generate_candidates(
             for prediction in predictions
             if cls.is_prediction_valid(product, prediction.value_tag)  # type: ignore
         ]
+        for candidate in candidates:
+            if candidate.value_tag:
+                # we normalize `value_tag` to the canonical taxonomy value
+                # it helps to deal with internationalization
+                # (`fr:sans-gluten` and `en:no-gluten` is the same tag)
+                value_tag = match_taxonomized_value(
+                    candidate.value_tag, TaxonomyType.label.name
+                )
+                if value_tag is not None:
+                    candidate.value_tag = value_tag
+
         taxonomy = get_taxonomy(InsightType.label.name)
         for candidate in select_deepest_taxonomized_candidates(candidates, taxonomy):
             insight = ProductInsight(**candidate.to_dict())

tests/unit/insights/test_importer.py

@@ -776,6 +776,18 @@ def test_is_parent_label(self, label, to_check_labels, expected, mocker):
                 Product({"code": DEFAULT_BARCODE, "labels_tags": ["en:vegan"]}),
                 [("en:ecoveg", False), ("en:max-havelaar", True)],
             ),
+            (
+                # fr:sans-gluten should be normalized into en:no-gluten
+                [
+                    Prediction(
+                        PredictionType.label,
+                        value_tag="fr:sans-gluten",
+                        automatic_processing=True,
+                    ),
+                ],
+                Product({"code": DEFAULT_BARCODE}),
+                [("en:no-gluten", True)],
+            ),
         ],
     )
     def test_generate_candidates(self, predictions, product, expected, mocker):