How to use hugging face in Rasa

Rasa Open Source
1

HI, I am a beginner in Rasa opensource. I am trying to build a banking chatbot using handle simple activities like balance check, fund transfer, account statement, FAQs. I want to use hugging face to extract custom entities such as account numbers, transfer amounts, card numbers.

I tried watching tutorials on hugging face on youtube, but almost all of them were few years old and wasn’t on Rasa 3.X. Can you please help me or guide me towards the correct direction, so I could find a proper way to use hugging face on my bot.

Thank you.

2

Hello! To integrate Hugging Face for custom entity extraction in Rasa 3.X, install the transformers library, create a custom entity extractor using Hugging Face models, train your Rasa model with this extractor, and integrate it into your Rasa pipeline. Refer to the Rasa documentation and check out the Hugging Face tutorial for detailed steps.

3

Hello, To integrate Hugging Face with Rasa 3.X for your banking chatbot, you can follow these steps:

  1. Install Hugging Face Transformers: Ensure you have the Hugging Face Transformers library installed in your environment.
  2. Create a Custom NLU Pipeline: Define a custom NLU pipeline in your config.yml file that includes a component for Hugging Face models.
  3. Train Your Model: Train your model using the custom NLU pipeline to recognize entities like account numbers and transfer amounts.
  4. Test and Iterate: Test your chatbot and refine the model as needed.

For detailed guidance, you can refer to the Rasa documentation on setting up LLMs and check out this tutorial on integrating Hugging Face with Rasa.

Best Regards, Jennifer Block