Entities with punctuation and space are not recognized

[jhzape (Jin)]

Hello,

I have an intent which contains space and punctuations as entity.

During training, there is user warning saying " Issue found while processing ‘’: Intent ‘input_UID’ has no examples." And during testing, the entity value cannot be correctly recognized. Can you please help? @koaning

[koaning (Vincent D. Warmerdam)]

Could you share the traceback of the warning? If you’re running this locally you should see it appear if you run rasa train from the command line. Also, could you share the Rasa version via rasa --version?

[jhzape (Jin)]

Thanks for the reply. Here attached the user warning together with version record.

Besides, I saw an answer to a similar question where Spacy tokenizer should replace white space. Therefore, I would like to also attach config file.config.yml (1.3 KB)

[koaning (Vincent D. Warmerdam)]

In the future, please attach any tracebacks/code examples as markdown in the text. That way it’s possible for me to copy and paste the code. Doing it with screenshots involves a lot of manual work on our end, especially because the yellow text doens’t contrast nicely with a white background.

I think I’ve found an issue. You’re using the RegexFeaturizer. This attaches machine learning features but it doesn’t directly detect entities. What you’re probably interested in instead is the RegexEntityExtractor component. There’s a youtube video here where you can see an example of the RegexEntityExtractor being used in a pipeline.

The main issue you see from Rasa X though is likely due to the fact that you have intents with no examples (as you can confirm from the yellow logs).

[jhzape (Jin)]

@koaning I appreciate your suggestion on how to post code formally. Besides, RegexEntityExtractor is exactly what I want and the UID can be correctly extracted now. I know you are an expert in customizing pipeline from both advanced workshop and forum. Thanks a lot.