How can I make Entity Extraction Case insensitive

I am working on a chatbot. My goal is to make our bot to be case insensitive. I am using rasa 2.8.1 version. The entities extracted are case sensitive. I also added case_sensitive: False to WhitespaceTokenizer. And later experiment with adding case_sensitive: False to RegexFeaturizer also. But it seems to be not working in case of entity extraction.

For example , I type the word “android” and our entity extractor identifies it properly as download_platform:android. {“entity”: “download_platform”, “value”: “android”} But When I type “Android” or “anDroiD”, the entity extractor recognize the word as download_platform: Android and download_platform: anDroiD respectively.

Is there any way to make the entity extraction case insensitive?

You can do this will synonyms and lookup tables or you can clean-up the user response in a form field validate method.

Thank you so much for the reply. adding synonyms will be fine for making the first letter capital.

- synonym: android
  examples: |
        - Android

But, if a user type capital letters in the middle with correct spelling adding all the possible ways is difficult. We are doing this for a banking domain and it has so many products in each section. I will check up with clean-up the user response in a form field validate method which I haven’t experimented.

Any other suggestions ?

If you’re using case_sensitive: False you don’t have to do every combination and the case used in the synonym doesn’t matter.

Any case combination for android will be changed to android in your synonym example.

1 Like

Thank you so much. It woks now.