Intent Recognition
Definition
Intent recognition assigns a linguistic query to a supported task.
Intent recognition here refers to assigning an utterance to a task class in a dialogue system. The system must also handle queries that do not belong to any of its supported classes.
Simply explained
As at reception, your utterance is assigned to a suitable task; hidden thoughts are not read in the process.
Why it matters in practice
Before deployment, supported intents and handling of unknown or unclear queries should be defined.
Example from business
Possible example: A service assistant distinguishes between appointment requests and status inquiries and forwards inappropriate requests for manual clarification.
Limits and pitfalls
- Not every query belongs to a known class. The cited study shows that good results on known intents do not automatically mean good detection of unknown intents.
Origin & history
Larson and co-authors published a dataset in 2019 for joint investigation of intent classification and out-of-scope queries.
Use in mid-sized companies
- Possible application areas are internal IT requests and appointment services; unknown intents would be handled via a designated fallback route.
Frequently asked questions
Does the system recognize thoughts?
No, linguistic queries are classified here.
Can every query simply be assigned to the most similar class?
That would be problematic if it lies outside the supported scope.
Sources
- An Evaluation Dataset for Intent Classification and Out-of-Scope Prediction · Larson et al. (2019): An Evaluation Dataset for Intent Classification and Out-of-Scope Prediction
Editorially reviewed on: