Joint Intent Detection and Slot Filling via CNN-LSTM-CRF

Abstract

Intent detection and slot filling are two main tasks in the domain of Spoken Language Understanding (SLU). The methods employed may treat the intent detection and slot filling as two independent tasks or use a joint model. Using a joint model takes into account the cross impact between the two tasks. In this article, we introduce CoBiC a new model combining CNN (Convolutional Neural Network), Bidirectional LSTM (Long Short-Term Memory) and CRF (Conditional Random Field) to extract the intents and the related slots. The same architecture of CoBiC can either be used as an independent model or joint model for intent detection and slot filling. Our method improves the state-of-the-art results on ATIS (Airline Travel Information Systems) benchmark. We also apply our model on a private dataset consisting of clients requests to a vocal assistant. The results demonstrate that CoBiC has strong generalization capability.