As a legal team, we have lots of policy PDF’s and we want to train Rasa Bot from these PDF instead of converting them in FaQ’s format due to bandwidth constraints.
Do we have a way to train the model from PDF which is in free text?
I’ve worked on this type of project and in the end converted to FAQ’s. There’s an interesting demo of the Google Univsersal Sentence Encoder that searches books. I tried this approach with Rasa but found that it didn’t address the user requirements.