Generative Encoder-Decoder Models for Task-Oriented Spoken Dialog Systems with Chatting Capability

Abstract

Generative encoder-decoder models offer great promise in developing domaingeneral dialog systems. However, they have mainly been applied to open-domain conversations.This paper presents a practical and novel framework for building task-oriented dialog systems based on encoder-decoder models. This framework enables encoder-decoder models to accomplish slot-value independent decision-making and interact with external databases. Moreover, this paper shows the flexibility of the proposed method by interleaving chatting capability with a slotfilling system for better out-of-domain recovery. The models were trained on both real-user data from a bus information system and human-human chat data. Results show that the proposed framework achieves good performance in both offline evaluation metrics and in task success rate with human users.

Hierarchical Attention and Knowledge Matching Networks With Information Enhancement for End-to-End Task-Oriented Dialog Systems

“…Besides database interaction, the generation procedure was also considered by researchers. The task-oriented dialog systems were equipped with chatting capability by inserting chats in original data [8]. However, the representations of dialog history are quite straightforward and rarely discussed in the models above.…”

“…It only needs text transcripts of dialogs without intermediate labels or handcrafted rules and can be adapted to different domains more easily than the first approach. Among all the end-to-end models, Seq2seq shows its potential in dialog systems in recent research [8], [9]. Seq2seq is an encoder-decoder model which is based on recurrent neural networks (RNN) and was invented for machine translation (MT) [10].…”

“…Seq2seq is an encoder-decoder model which is based on recurrent neural networks (RNN) and was invented for machine translation (MT) [10]. It has been widely used in open domain conversations [11]- [15] and end-to-end taskoriented DS [8], [9], [16], [17].…”

A Working Memory Model for Task-oriented Dialog Response Generation

“…Task-oriented dialog systems, such as hotel booking or technical support service, help users to achieve specific goals with natural language. Compared with traditional pipeline solutions (Williams and Young, 2007;Young et al, 2013;Wen et al, 2017), end-to-end approaches recently gain much attention (Zhao et al, 2017;Eric and Manning, 2017a;Lei et al, 2018), because they directly map dialog history to the output responses and consequently reduce human effort for modular designs and hand-crafted state labels. To effectively incorporate KB information and perform knowledge-based reasoning, memory augmented models have been proposed (Bordes et al, 2017;Seo et al, 2017;Eric and Manning, 2017b;Madotto et al, 2018;Raghu et al, 2018;Reddy et al, 2019;Wu et al, 2019).…”

Multi-Granularity Representations of Dialog

“…Constructing meaningful representations of dialog is challenging. To effectively represent the dialog context, a latent dialog representation must contain the information necessary to (1) estimate a belief state over user goals (Williams et al, 2013), (2) track entity mentions (Zhao et al, 2017), (3) resolve anaphora co-references (Mitkov, 2014), (4) model the communicative purpose of an utterance (Core and Allen, 1997) and (5) resolve ambiguity in natural language. A large focus area of dialog research is the development of neural architectures which learn effective represen-tations of the input Zhou et al, 2018).…”