CDS: Contextual Spoken Language Understanding Model Based on Dual Semantics. (November 2020)

• Record Type: Journal Article
• Authors: Ma, Kun; Hu, Minghao; Li, Fangzhao; Luo, Li
• Is Part Of: Journal of physics. Volume 1651(2020)
• Publication Date: 2020-11
• Subjects: Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/1651/1/012051
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: Spoken language understanding (SLU) in multi-turn task-oriented dialogue systems usually needs to take historical contextual information into account. To achieve this, current approaches encode dialogue utterances with contextual information to improve the accuracy of slot filling and intent detection. However, this kind of approach ignores the different influences of contextual information on dialogue utterances. Some utterances with independent context will be misunderstood due to the addition of noisy contextual information. In this paper, we propose a Dual Semantics model(CSD) that is based on dual semantics. Each dialogue utterance is encoded based on independent semantics and contextual semantics, where the contributions of these two semantic information are dynamically decided. A large number of experiments on two datasets verify the effectiveness of our approach. Specifically, the accuracy of slot filling at the KVRET dataset increases by 3.3% compared to the baseline model. A dataset of psychological service appointments(PSA), is further collected for the verification of our approach. Compared with existing methods, the accuracy is improved by 4.9% on this dataset.