How to Make a RASA Chatbot Learn from a PDF Without Defining Intents Manually?

Hello everyone, I want to build a chatbot using RASA that should learn from a PDF document and retrieve information based on its content.

What I Want to Achieve:

The chatbot should function as a customer support assistant by searching for relevant information in the PDF. However, I do not want to manually define intents and rules for every possible query. Instead, I want the chatbot to automatically learn from the PDF and generate responses based on its content—similar to how ChatGPT would process and answer questions from a document.

Question:

Is there a way to implement this in RASA? How can I make the chatbot extract information directly from the PDF and respond to user queries accordingly without manually specifying intents and rules? Any guidance or example implementations would be greatly appreciated!

Hi @iggy, it’s great that you are building with Rasa. It would be much easier to implement this functionality using Rasa Pro & CALM. CALM has EnterpriseSearchPolicy which handles all how-to questions and can extract information from provided docs. You can get free developer key here. Read more about Enterpise Search Policy here. Let me know if you have any further questions.