Goal-Embedded Dual Hierarchical Model for Task-Oriented Dialogue Generation

Proceedings of the 23rd Conference on Computational Natural Language Learning, pages 798–811 Hong Kong, China, November 3-4, 2019. c⃝2019 Association for Computational Linguistics 798 Goal-Embedded Dual Hierarchical Model for Task-Oriented Dialogue Generation Yi-An Lai Amazon ML Seattle yianl@amazon.com Arshit Gupta Amazon ML Seattle arshig@amazon.com Yi Zhang Amazon ML Seattle yizhngn@amazon.com Abstract Hierarchical neural networks are often used to model inherent structures within dialogues. For goal-oriented dialogues, these models miss a mechanism adhering to the goals and neglect the distinct conversational patterns be- tween two interlocutors. In this work, we pro- pose Goal-Embedded Dual Hierarchical At- tentional Encoder-Decoder (G-DuHA) able to center around goals and capture interlocutor- level disparity while modeling goal-oriented dialogues. Experiments on dialogue genera- tion, response generation, and human evalu- ations demonstrate that the proposed model successfully generates higher-quality, more di- verse and goal-centric dialogues. Moreover, we apply data augmentation via goal-oriented dialogue generation for task-oriented dialog systems with better performance achieved. 1 Introduction Modeling a probability distribution over word se- quences is a core topic in natural language pro- cessing, with language modeling being a flagship problem and mostly tackled via recurrent neu- ral networks (RNNs) (Mikolov and Zweig, 2012; Melis et al., 2017; Merity et al., 2018). Recently, dialogue modeling has drawn much attention with applications to response generation (Serban et al., 2016a; Li et al., 2016b; Asghar et al., 2018) or data augmentation (Yoo et al., 2019). It’s inherently different from language modeling as the conversation is conducted in a turn-by-turn nature. (Serban et al., 2016b) im- poses a hierarchical structure on encoder-decoder to model this utterance-level and dialogue-level structures, followed by (Serban et al., 2016c; Chen et al., 2018; Le et al., 2018a). However, when modeling dialogues involving two interlocutors center around one or more goals, these systems generate utterances with the great- est likelihood but without a mechanism sticking to Hi , I need a train to Boston. Sure, what day will you be traveling ? I need to leave on tomorrow and arrive before 5 pm. Train 3315 leaves at 10 am, would you like to book it? No, I only need the price. The price is $100. I also need a hotel to stay. Sure. Do you have a price range in mind? … Yes, please book 3 tickets. The reference number is RVUSFG. Is there anything else I can help? I’m also looking for information on a museum. … Goals: (1) Domain: Train, Requests: [Price], Book: False (2) Domain: Hotel, Requests: [Area, Phone], Book: True On-Goal Dialogue Off-Goal Dialogue Figure 1: On-goal dialogues follow the given goals such as no booking of train tickets and a hotel reserva- tion. Off-goal dialogues have context switches to other domains non-relevant to goals. the goals. This makes them go off the rails and fail to model context-switching of goals. Most of the generated conversations become off-goal dia- logues with utterances being non-relevant or con- tradicted to goals rather than on-goal dialogues. The differences are illustrated in Figure 1. Besides, two interlocutors in a …

Content truncated to 500 words