In this paper, a deep reinforcement learning(DRL) based multimodal coaching model (DCM) for slot filling in SLU is proposed. This new model functions as a coach to help an RNN based tagger to learn the wrong labeled slots, hence may further improve the performance of an SLU system. Besides, users can also coach the model by correcting its mistakes and help it progress further. The performance of DCM is evaluated on two datasets: one is the benchmark ATIS corpus dataset, another is our in-house dataset with three different domains. It shows that the new system gives a better performance than the current state-of-the-art results on ATIS by using DCM. Furthermore, we build a demo app to further explain how user's input can also be used as a real-time coach to improve model's performance even more.
Cite as: Wang, Y., Patel, A., Shen, Y., Jin, H. (2018) A Deep Reinforcement Learning Based Multimodal Coaching Model (DCM) for Slot Filling in Spoken Language Understanding(SLU). Proc. Interspeech 2018, 3444-3448, doi: 10.21437/Interspeech.2018-1379
@inproceedings{wang18e_interspeech,
author={Yu Wang and Abhishek Patel and Yilin Shen and Hongxia Jin},
title={{A Deep Reinforcement Learning Based Multimodal Coaching Model (DCM) for Slot Filling in Spoken Language Understanding(SLU)}},
year=2018,
booktitle={Proc. Interspeech 2018},
pages={3444--3448},
doi={10.21437/Interspeech.2018-1379}
}