Intent Classification
Labeling the underlying goal behind a user's text or voice request.
What this means for trainers
Pay attention to what the user is actually trying to accomplish, not the keywords they used — literal keyword matching is the most common mistake in this task type.
Intent classification assigns a request to the goal behind it rather than its literal wording. Two very differently worded requests, like asking about the weather this weekend and asking whether an umbrella will be needed Saturday, might both map to the same underlying intent despite sharing almost no words in common.
This task underpins conversational assistants and voice systems. Once intent is identified, a separate slot-filling step extracts the specific parameters, such as a date or a location, needed to fulfill the request.
The hard part of intent classification is usually taxonomy design rather than the labeling itself. Overlapping or ambiguous intent categories create the same kind of ambiguity problems seen in any other classification task, and a taxonomy with too many near-duplicate intents makes consistent labeling almost impossible regardless of how careful the annotator is.
Related terms
Put this into practice
Browse open data annotation and evaluation roles from Mercor, Micro1, Outlier, and more.
Browse AI training jobs