I am new to chatbots and Rasa looks amazing. That being said, if I want to create a simple bot where text messages are merely classified as offensive or not is Rasa overkill? The bot wouldn’t really have an interactive chat (intent) component to it unless it was for configuration.
Hi @jeffnb! This does not sound like a bot to me. It sounds like you want a machine learning model to classify if a message is offensive or not. You definitely won’t need a lot of what Rasa offers, but it should be much easier to achieve this using Rasa than trying to do it from scratch.
The most challenging part will be creating a good dataset of offensive and not offensive messages. I recommend collecting real messages from users using Rasa X, taking advantage of the NLU Inbox to mark these messages as correctly classified or not, versioning your training data and pushing it to a remote Git repo using Integrated Version Control, and then running rasa test nlu in an automated fashion to ensure your model improves as you add more data (e.g. like this GitHub action). As you start to collect and label more data, you can retrain and deploy the improved model. As you do this more and more, you should start to click
mark as correct more and more often