SAS: Dialogue State Tracking via Slot Attention and Slot Information Sharing

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6366–6375 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 6366 SAS: Dialogue State Tracking via Slot Attention and Slot Information Sharing Jiaying Hu1, Yan Yang1,3∗, Chengcai Chen2, Liang He1,3, Zhou Yu4 1East China Normal University 2Xiaoi Research, Xiaoi Robot Technology Co., Ltd 3Shanghai Key Laboratory of Multidimensional Information Processing 4University of California, Davis 51184506058@stu.ecnu.edu.cn, {yanyang,lhe}@cs.ecnu.edu.cn arlenecc@xiaoi.com, joyu@ucdavis.edu Abstract Dialogue state tracker is responsible for in- ferring user intentions through dialogue his- tory. Previous methods have difficulties in handling dialogues with long interaction con- text, due to the excessive information. We propose a Dialogue State Tracker with Slot Attention and Slot Information Sharing (SAS) to reduce redundant information’s interference and improve long dialogue context tracking. Specially, we first apply a Slot Attention to learn a set of slot-specific features from the original dialogue and then integrate them us- ing a Slot Information Sharing. The sharing improve the models ability to deduce value from related slots. Our model yields a sig- nificantly improved performance compared to previous state-of-the-art models on the Multi- WOZ dataset. 1 Introduction The recent global adoption of personal assistants such as Alexa and Siri made dialogue system a more popular topic in research. The major dif- ference between dialogue systems and question- answering is that dialogue systems need to track dialogue history effectively. So, we normally use a dialogue state tracking component to track user’s intention throughout the conversation. A dialogue state is typically composed as a set of slot value pairs in a task-oriented dialogue, such as “hotel- internet-yes”. It means the slot “hotel-internet” has a value of “yes”. Early dialogue state tracking model needs a pre- defined ontology which means the values of ev- ery slot are enumerated in advance (Henderson et al., 2014; Mrkˇsi´c et al., 2017; Zhong et al., 2018; Sharma et al., 2019). Such practice is inefficient and costly. The large number of possible slot-value pairs makes deploying these models in the real-life ∗*Corresponding author. applications difficult (Rastogi et al., 2017). This difficulty is further amplified in multi-domain dia- logue state tracking where the dialogues have more than one tasks. Because the manual effort grows exponentially with the complexity of the dialogues. In (Wu et al., 2019), Wu et al. introduced a transfer- able dialogue state generator (TRADE), which can generate dialogue states from utterances using a copy mechanism. This generative model achieved relative good performance, but it still has trouble in extracting relevant information from the original dialogues. For example, a user may tell the agent that he/she needs a taxi in a turn, but the taxi’s departure location is implicitly mentioned several turns ago. Inspired by the (Chen et al., 2017; Chen, 2018), (Chen et al., 2019) studied on utilizing atten- tion mechanism to deal with the long distance slot carryover problem. In their work, they first fused the information of the slot, its corresponding value and the dialogue distance into a single vector. Then they …

Content truncated to 500 words