When new messages get shown in the NLU-inbox of rasa-x it preselects an entity extracted by a custom entiy extractor. this causes it to be entered in the NLU which causes the DIET-clasifier to be trained on that entity.
I saw that the entities extracted by duckling do not get preselected in the NLU-Inbox of rasa-x and when you mark them as correct, only the text apears in the nlu without the entity annotation. Which is correct ofcourse.
Now why is this not possible for a custom entity extractor. We need a way to tell rasa-x that an entity should not be anotated in the nlu.
Below are some examples of my problem
Here we see that the text is annotated and that there are 2 entities detected, an postalcode and a number. both entities are extracted by a pretrained entitie extractor so rasa-x does not let me change anything.
Now in nlu.md we see that rasa-x annotated the postalcode but did nothing with the number entity. Is there a way I can tell rasa-x to stop annotating the postalcodes in the nlu?