Scalable and Accurate Dialogue State Tracking via Hierarchical Sequence Generation

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 1876–1885, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 1876 Scalable and Accurate Dialogue State Tracking via Hierarchical Sequence Generation Liliang Ren, Jianmo Ni and Julian McAuley University of California, San Diego La Jolla, CA92093 {lren,jin018,jmcauley}@ucsd.edu Abstract Existing approaches to dialogue state track- ing rely on pre-defined ontologies consisting of a set of all possible slot types and values. Though such approaches exhibit promising performance on single-domain benchmarks, they suffer from computational complexity that increases proportionally to the number of pre-defined slots that need tracking. This is- sue becomes more severe when it comes to multi-domain dialogues which include larger numbers of slots. In this paper, we investi- gate how to approach DST using a generation framework without the pre-defined ontology list. Given each turn of user utterance and system response, we directly generate a se- quence of belief states by applying a hierarchi- cal encoder-decoder structure. In this way, the computational complexity of our model will be a constant regardless of the number of pre- defined slots. Experiments on both the multi- domain and the single domain dialogue state tracking dataset show that our model not only scales easily with the increasing number of pre-defined domains and slots but also reaches the state-of-the-art performance. 1 Introduction A Dialogue State Tracker (DST) is a core compo- nent of a modular task-oriented dialogue system (Young et al., 2013). For each dialogue turn, a DST module takes a user utterance and the dia- logue history as input, and outputs a belief esti- mate of the dialogue state. Then a machine action is decided based on the dialogue state according to a dialogue policy module, after which a machine response is generated. Traditionally, a dialogue state consists of a set of requests and joint goals, both of which are rep- resented by a set of slot-value pairs (e.g. (request, phone), (area, north), (food, Japanese)) (Hender- son et al., 2014). In a recently proposed multi- DST Models ITC NBT-CNN (Mrksic et al., 2017) O(mn) MD-DST (Rastogi et al., 2017) O(n) GLAD (Zhong et al., 2018) O(mn) StateNet PSI (Ren et al., 2018) O(n) TRADE (Wu et al., 2019) O(n) HyST (Goel et al., 2019) O(n) DSTRead (Gao et al., 2019) O(n) Table 1: The Inference Time Complexity (ITC) of pre- vious DST models. The ITC is calculated based on how many times inference must be performed to complete a prediction of the belief state in a dialogue turn, where m is the number of values in a pre-defined ontology list and n is the number of slots. domain dialogue state tracking dataset, MultiWoZ (Budzianowski et al., 2018), a representation of dialogue state consists of a hierarchical structure of domain, slot, and value is proposed. This is a more practical scenario since dialogues often in- clude multiple domains simultaneously. Many recently proposed DSTs (Zhong et al., 2018; Ramadan et …

Content truncated to 500 words