I’ve made this app that can cluster short texts.
It’s something I’m still working on but this is useful to some Rasa users. When you’re just starting out making a virtual assistant then you might have a lot of customer feedback but you don’t know what main themes you can distill and add as intents for an assistant. The goal of this app is to make it just slightly easier to start exploring this.
A good step two is to also allow for labeling. I’ve created this tool called human-learn a while ago that makes this sort of thing easy in jupyter. I’m looking forward to get something like this also working in streamlit. Just need to make sure that I handle the “state” part properly because when you start labeling… there’s a lot of it .