Hierarchical Dialogue Understanding with Special Tokens and Turn-level Attention

TL;DR

The paper introduces HiDialog, a hierarchical dialogue understanding model that uses special tokens and turn-level attention to effectively capture semantic dynamics, achieving state-of-the-art results across multiple dialogue understanding tasks.

Contribution

It proposes a novel hierarchical approach with special tokens and turn-level attention, enhancing dialogue understanding performance.

Findings

Achieves state-of-the-art results on dialogue relation extraction.

Outperforms existing methods in emotion recognition and act classification.

Demonstrates effectiveness of hierarchical modeling with special tokens.

Abstract

Compared with standard text, understanding dialogue is more challenging for machines as the dynamic and unexpected semantic changes in each turn. To model such inconsistent semantics, we propose a simple but effective Hierarchical Dialogue Understanding model, HiDialog. Specifically, we first insert multiple special tokens into a dialogue and propose the turn-level attention to learn turn embeddings hierarchically. Then, a heterogeneous graph module is leveraged to polish the learned embeddings. We evaluate our model on various dialogue understanding tasks including dialogue relation extraction, dialogue emotion recognition, and dialogue act classification. Results show that our simple approach achieves state-of-the-art performance on all three tasks above. All our source code is publicly available at https://github.com/ShawX825/HiDialog.