Issue while classifying intent

Misaligned entity annotation for 'The Adventures of Robin Hood (1938' in sentence 'is The Adventures of Robin Hood (1938) is running in theater on riffa' with intent 'movie_by_user+areazz_chosenby_user'. Make sure the start and end values of the annotated training examples end at token

here in the above example “)” is getting skiped like its not taking any special character ,any kind of suggestions wll be helpful

The second part of the warning should continue like this

Make sure the start and end values of the annotated training examples end at token boundaries (e.g. don't include trailing whitespaces or punctuation).

Did you check your training data? How does the example look like? Does it actually contains an entity?

Also, it is just a warning, so the one entity will be ignored, but you should nevertheless be able to train your model. Is that the case? Or do you get any other error along the way?

yes @Tanja i checked actually its not considering special character like it’s skipping the last special character present in last word example “raaz 3 (3D)” it’s returning me as “raaz 3 (3D”

What pipeline are you using? And what Rasa Version do you use?

I am using supervised embeddings as you can se below language: en pipeline: supervised_embeddings policies:

  • name: MemoizationPolicy
  • name: KerasPolicy
  • name: MappingPolicy

I just checked the code. During preprocessing of the text any punctuation characters are removed. So, raaz 3 (3D) becomes raaz 3 3D. Let’s say the entity raaz 3 (3D) starts at position 5 in the text and ends at position 16. Due to the fact that we removed the closing bracket ) during preprocessing the end position of the entity does not match anymore/is not present anymore. Thus, the entity is misaligned. To avoid such problems, you should avoid punctuation characters in your entities, especially at the end of your entities. If an entity is misaligned, it will be ignored, so you will have one training example less.