Merge pull request #6549 from RasaHQ/slow-crf-entity-extractor-training

Fix slow training of CRFEntityExtractor when using Entity Roles and Groups
RasaHQ · Sep 3, 2020 · c7f7d37 · c7f7d37
2 parents f07b286 + bb97548
commit c7f7d37
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 4 deletions.
diff --git a/changelog/6549.bugfix.rst b/changelog/6549.bugfix.rst
@@ -0,0 +1 @@
+Fix slow training of ``CRFEntityExtractor`` when using Entity Roles and Groups.
diff --git a/rasa/nlu/extractors/crf_entity_extractor.py b/rasa/nlu/extractors/crf_entity_extractor.py
@@ -406,14 +406,18 @@ def _create_features_for_token(
                 # get the features to extract for the token we are currently looking at
                 current_feature_idx = pointer_position + half_window_size
                 features = configured_features[current_feature_idx]
+
+                prefix = prefixes[current_feature_idx]
+
                 # we add the 'entity' feature to include the entity type as features
                 # for the role and group CRFs
+                # (do not modify features, otherwise we will end up adding 'entity'
+                # over and over again, making training very slow)
+                additional_features = []
                 if include_tag_features:
-                    features.append("entity")
-
-                prefix = prefixes[current_feature_idx]
+                    additional_features.append("entity")
 
-                for feature in features:
+                for feature in features + additional_features:
                     if feature == "pattern":
                         # add all regexes extracted from the 'RegexFeaturizer' as a
                         # feature: 'pattern_name' is the name of the pattern the user