Evaluating and Improving Context Attention Distribution on Multi-Turn Response Generation using Self-Contained Distractions

TL;DR

This paper introduces a new metric called DAS ratio for evaluating how well multi-turn dialogue models distribute attention across context, and proposes an optimization method using self-contained distractions to enhance this aspect.

Contribution

It presents a novel attention-based metric for multi-turn response generation and an optimization strategy that improves context attention distribution in dialogue models.

Findings

Models with similar perplexity differ in attention distribution quality.

The proposed optimization improves models' attention distribution by about 10%.

Enhanced attention distribution correlates with better multi-turn response quality.

Abstract

Despite the rapid progress of open-domain generation-based conversational agents, most deployed systems treat dialogue contexts as single-turns, while systems dealing with multi-turn contexts are less studied. There is a lack of a reliable metric for evaluating multi-turn modelling, as well as an effective solution for improving it. In this paper, we focus on an essential component of multi-turn generation-based conversational agents: context attention distribution, i.e. how systems distribute their attention on dialogue's context. For evaluation of this component, We introduce a novel attention-mechanism-based metric: DAS ratio. To improve performance on this component, we propose an optimization strategy that employs self-contained distractions. Our experiments on the Ubuntu chatlogs dataset show that models with comparable perplexity can be distinguished by their ability on context attention distribution. Our proposed optimization strategy improves both non-hierarchical and hierarchical models on the proposed metric by about 10% from baselines.