Team Flow at DRC2022: Pipeline System for Travel Destination Recommendation Task in Spoken Dialogue

TL;DR

This paper describes a dialogue system developed for the DRC2022 competition, utilizing GPT-2 based NLU and NLG modules with rule-based DST and policy, highlighting challenges in training data diversity and policy effectiveness.

Contribution

The paper presents a pipeline dialogue system with GPT-2 components and rule-based modules, applied in a competitive setting to explore system limitations and improvements.

Findings

Limited training data affected NLU performance.

Policy design impacted recommendation success.

System performance was constrained by data and policy issues.

Abstract

To improve the interactive capabilities of a dialogue system, e.g., to adapt to different customers, the Dialogue Robot Competition (DRC2022) was held. As one of the teams, we built a dialogue system with a pipeline structure containing four modules. The natural language understanding (NLU) and natural language generation (NLG) modules were GPT-2 based models, and the dialogue state tracking (DST) and policy modules were designed on the basis of hand-crafted rules. After the preliminary round of the competition, we found that the low variation in training examples for the NLU and failed recommendation due to the policy used were probably the main reasons for the limited performance of the system.