Multi intents behind the scene

Hi there, could someone explain to me what goes on in DIET when we use multi intent labels? Are they considered as standalone intents, do the individual intents somehow feed into the prediction or the other way around. Is it useful to just concatenate the phrases from the individual intents to use as examples for the multi intent or is that just redundant?

A lot of questions… But hopefully a simple answer!


Hi Ifabing,

could you clarify what you mean with “when we use multi intent labels”? Are you using the standard files to train the model or something via Jupyter notebook? In general Rasa offers NLU models that perform classification of intents, as in, one intent is classified.