IDALC: A Semi-Supervised Framework for Intent Detection and Active Learning based Correction

Researchers introduce IDALC, a semi-supervised framework for voice-controlled dialog systems that improves intent detection and reduces manual annotation costs. The system achieves 5-10% higher accuracy and 4-8% better macro-F1 scores while requiring annotation of only 6-10% of unlabeled data.

• →IDALC framework addresses the problem of system-rejected utterances in voice-controlled dialog systems that require costly manual annotation.
• →The semi-supervised approach achieves 5-10% higher accuracy and 4-8% improvement in macro-F1 scores compared to baseline methods.
• →The system reduces annotation costs significantly, requiring human labeling for only 6-10% of available unlabeled data.
• →Voice-controlled dialog systems often reject utterances with low confidence even for known intents, creating efficiency bottlenecks.
• →The framework enables dynamic retraining with new intents from previously rejected queries without extensive manual labeling.