Scalable multi-domain dialogue state tracking

By Abhinav Rastogi, Dilek Z. Hakkani-Tür, Larry Heck. 2017

115 citations recorded by the source corpus.

Dialogue state tracking (DST) is a key component of task-oriented dialogue systems. DST estimates the user's goal at each user turn given the interaction until then. State of the art approaches for state tracking rely on deep learning methods, and represent dialogue state as a distribution over all possible slot values for each slot present in the ontology. Such a representation is not scalable when the set of possible values are unbounded (e.g., date, time or location) or dynamic (e.g., movies or usernames). Furthermore, training of such models requires labeled data, where each user turn is annotated with the dialogue state, which makes building models for new domains challenging. In this paper, we present a scalable multi-domain deep learning based approach for DST. We introduce a novel framework for state tracking which is independent of the slot value set, and represent the dialogue state as a distribution over a set of values of interest (candidate set) derived from the dialogue history or knowledge. Restricting these candidate sets to be bounded in size addresses the problem of slot-scalability. Furthermore, by leveraging the slot-independent architecture and transfer learning, we show that our proposed approach facilitates quick adaptation to new domains.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Natural Language Processing

Citations in the corpus

A Simple Language Model for Task-Oriented Dialogue (2020, background): “These components include understanding user intent via intent detection [26], tracking the constraints imposed by the user via dialogue state tracking [20, 32, 41, 34, 53, 59, 64, 8, 19], determining system actions via dialogue policy [51], and using dedicated response generation components [49].”
MultiWOZ - A Large-Scale Multi-Domain Wizard-of-Oz Dataset for Task-Oriented Dialogue Modelling (2018, background): “Since multi-domain dialogue state tracking is still in its infancy and there are not many comparable approaches available (Rastogi et al., 2017), we instead report our state-of-the-art result on the restaurant subset of the MultiWOZ corpus as the reference baseline. proposed method (Ramadan et al.,…”
Towards Scalable Multi-domain Conversational Agents: The Schema-Guided Dialogue Dataset (2019, background): “As virtual assistants incorporate diverse domains, recent work has focused on zero-shot modeling (Bapna et al. 2017; Xia et al. 2018; Shah et al. 2019), domain adaptation and transfer learning techniques (Yang, Salakhutdinov, and Co-hen 2017; Rastogi, Hakkani-T¨ur, and Heck 2017; Zhu and Yu 2018).”
Transferable Multi-Domain State Generator for Task-Oriented Dialogue Systems (2019, background): “Lastly, several approaches have been proposed for continual learning in the machine learning community (Kirkpatrick et al., 2017; Lopez-Paz et al., 2017; Rusu et al., 2016; Fernando et al., 2017; Lee et al., 2017), especially in image recognition tasks (Aljundi et al., 2017; Rannen et al., 2017).”
Multi-domain Dialogue State Tracking as Dynamic Knowledge Graph Enhanced Question Answering (2019, background): “Models that handles scalable multi-domain DST have then been proposed (Ramadan et al., 2018; Rastogi et al., 2017).”
Toward Scalable Neural Dialogue State Tracking Model (2018, methodology): “Therefor, our proposed model is simpler since it contains only one conditioned encoder for user utterance, whereas Rastogi et al. (2017) model requires two independet conditioned encoder.”
MultiWOZ 2.2 : A Dialogue Dataset with Additional Annotation Corrections and State Tracking Baselines (2020, background): “…the locations in the user and system utterances where slot values have been mentioned, to make the system efficient and generalizable to new slot values (Rastogi et al., 2017; Wu et al., 2019; Zhang et al., 2019; Rastogi et al., 2019; Xu and Hu, 2018; Zhou and Small, 2019; Gao et al., 2019).”
An End-to-end Approach for Handling Unknown Slot Values in Dialogue State Tracking (2018, methodology): “but both depend on candidate lists as given and hence are not designed to handle unknown (different from unseen) slot values; The scalable DST model (Rastogi et al., 2018), although addressing the same problem of unbounded value set, re-”
Multi-task Learning for Joint Language Understanding and Dialogue State Tracking (2018, methodology): “Rastogi et al. 2017 proposed the use of candidate sets in DST for ef-ﬁciently handling slots with a large set of values.”
Dialog State Tracking: A Neural Reading Comprehension Approach (2019, unknown): “To address these concerns, a class of methods employing scoring mechanisms to predict the slot value from a endogenously deﬁned set of candidates have been proposed (Rastogi et al., 2017; Goel et al., 2018).”
Data Augmentation for Copy-Mechanism in Dialogue State Tracking (2020, unknown): “The Pointer Network is designed to learn the conditional probability of an output sequence with elements that are discrete tokens corresponding to positions in an input sequence.”
Neural Assistant: Joint Action Prediction, Response Generation, and Latent Knowledge Reasoning (2019, unknown): “Deep learning has been applied to task-oriented dialog in many recent studies [4, 22, 25, 13, 11, 23, 1].”
Building a Conversational Agent Overnight with Dialogue Self-Play (2018, background): “…for training dialogue models from recent literature: either component-wise models for language understanding (Bapna et al. (2017)), state tracking (Rastogi et al. (2017)), dialogue policy (Shah et al. (2016)) and language generation (Nayak et al. (2017)), or end-to-end models (Wen et al. (2016)).”
Dialogue State Tracking with a Language Model using Schema-Driven Prompting (2021, background): “(Rastogi et al., 2017) should be considered as a kind of oracle upper bound performance because the target slot value is guaranteed to be in the candidate list and consider by the model.”