We’re trying to build a chatbot for mental health. Planning to train the bot to understand the user’s feeling using NER. Which is the easiest way to do this?
When you’re trying to detect “the feeling” … is that more like an intent or more like an entity? It sounds like you’re more interested in detecting a sentiment while the entity is typically used to detect a substring of interest. Things like dates, times, emailadresses, etc.
Ok understood. We wanted to keep a track of the user’s emotions, as they are using the bot everyday, that’s why I tried using entities. Just wanted to ask, is there any pretrained NER model (like spacy) for that?
There are pretrained sentiment models out there, but I must stress, these are not entity models. They are more like classifiers, which, Rasa does not natively support. You’d have to write your own custom component for that.
Got it. Thanks for helping.