Do Response Selection Models Really Know What's Next? Utterance Manipulation Strategies for Multi-turn Response Selection

TL;DR

This paper introduces utterance manipulation strategies (UMS) to improve multi-turn response selection models by emphasizing dialog coherence, demonstrating significant performance gains across various datasets and languages.

Contribution

The paper proposes self-supervised utterance manipulation strategies that enhance dialog coherence in response selection models, addressing limitations of existing approaches that ignore temporal dependencies.

Findings

UMS significantly improve response selection accuracy.

Models trained with UMS outperform state-of-the-art on multiple benchmarks.

UMS are effective across different languages and models.

Abstract

In this paper, we study the task of selecting the optimal response given a user and system utterance history in retrieval-based multi-turn dialog systems. Recently, pre-trained language models (e.g., BERT, RoBERTa, and ELECTRA) showed significant improvements in various natural language processing tasks. This and similar response selection tasks can also be solved using such language models by formulating the tasks as dialog--response binary classification tasks. Although existing works using this approach successfully obtained state-of-the-art results, we observe that language models trained in this manner tend to make predictions based on the relatedness of history and candidates, ignoring the sequential nature of multi-turn dialog systems. This suggests that the response selection task alone is insufficient for learning temporal dependencies between utterances. To this end, we propose utterance manipulation strategies (UMS) to address this problem. Specifically, UMS consist of several strategies (i.e., insertion, deletion, and search), which aid the response selection model towards maintaining dialog coherence. Further, UMS are self-supervised methods that do not require additional annotation and thus can be easily incorporated into existing approaches. Extensive evaluation across multiple languages and models shows that UMS are highly effective in teaching dialog consistency, which leads to models pushing the state-of-the-art with significant margins on multiple public benchmark datasets.