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

TL;DR

This paper introduces G-DuHA, a hierarchical neural model that emphasizes goals and interlocutor differences to generate more relevant, diverse, and goal-focused dialogues, improving task-oriented dialogue systems.

Contribution

The paper presents a novel goal-embedded dual hierarchical model that captures goal adherence and interlocutor disparities in dialogue generation.

Findings

Generated dialogues are more diverse and goal-centric.

Model outperforms baselines in dialogue quality and relevance.

Data augmentation with generated dialogues enhances task-oriented system performance.

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 between two interlocutors. In this work, we propose Goal-Embedded Dual Hierarchical Attentional Encoder-Decoder (G-DuHA) able to center around goals and capture interlocutor-level disparity while modeling goal-oriented dialogues. Experiments on dialogue generation, response generation, and human evaluations demonstrate that the proposed model successfully generates higher-quality, more diverse and goal-centric dialogues. Moreover, we apply data augmentation via goal-oriented dialogue generation for task-oriented dialog systems with better performance achieved.