Efficient Context and Schema Fusion Networks for Multi-Domain Dialogue State Tracking

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 766–781 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 766 Efficient Context and Schema Fusion Networks for Multi-Domain Dialogue State Tracking Su Zhu, Jieyu Li, Lu Chen∗and Kai Yu∗ State Key Laboratory of Media Convergence Production Technology and Systems MoE Key Lab of Artificial Intelligence, AI Institute, Shanghai Jiao Tong University SpeechLab, Department of Computer Science and Engineering Shanghai Jiao Tong University, Shanghai, China {paul2204,oracion,chenlusz,kai.yu}@sjtu.edu.cn Abstract Dialogue state tracking (DST) aims at estimat- ing the current dialogue state given all the pre- ceding conversation. For multi-domain DST, the data sparsity problem is a major obstacle due to increased numbers of state candidates and dialogue lengths. To encode the dialogue context efficiently, we utilize the previous di- alogue state (predicted) and the current dia- logue utterance as the input for DST. To con- sider relations among different domain-slots, the schema graph involving prior knowledge is exploited. In this paper, a novel context and schema fusion network is proposed to encode the dialogue context and schema graph by us- ing internal and external attention mechanisms. Experiment results show that our approach can outperform strong baselines, and the previous state-of-the-art method (SOM-DST) can also be improved by our proposed schema graph. 1 Introduction Dialogue state tracking (DST) is a key component in task-oriented dialogue systems which cover cer- tain narrow domains (e.g., booking hotel and travel planning). As a kind of context-aware language understanding task, DST aims to extract user goals or intents hidden in human-machine conversation and represent them as a compact dialogue state, i.e., a set of slots and their corresponding values. For example, as illustrated in Fig. 1, (slot, value) pairs like (name, huntingdon marriott hotel) are extracted from the dialogue. It is essential to build an accurate DST for dialogue management (Young et al., 2013), where dialogue state determines the next machine action and response. Recently, motivated by the tremendous growth of commercial dialogue systems like Apple Siri, Microsoft Cortana, Amazon Alexa, or Google As- sistant, multi-domain DST becomes crucial to help ∗The corresponding authors are Lu Chen and Kai Yu. Figure 1: An example of multi-domain dialogues. Ut- terances at the left side are from the system agent, and utterances at the right side are from a user. The dia- logue state of each domain is represented as a set of (slot, value) pairs. users across different domains (Budzianowski et al., 2018; Eric et al., 2019). As shown in Fig. the dialogue covers three domains (i.e., Hotel,1, Attraction and Taxi). The goal of multi- domain DST is to predict the value (including NONE) for each domain-slot pair based on all the preceding dialogue utterances. However, due to increasing numbers of dialogue turns and domain- slot pairs, the data sparsity problem becomes the main issue in this field. To tackle the above problem, we emphasize that DST models should support open-vocabulary based value decoding, encode context efficiently and in- corporate domain-slot relations: 1. Open-vocabulary DST is essential …

Content truncated to 500 words