Speaker Turn Modeling for Dialogue Act Classification

TL;DR

This paper introduces a method for dialogue act classification that incorporates speaker turn changes by learning conversation-invariant speaker turn embeddings, improving accuracy on benchmark datasets.

Contribution

It proposes a novel approach to model speaker turns in dialogue, enhancing dialogue act classification by integrating turn embeddings with utterance representations.

Findings

Model achieves superior performance on three benchmark datasets.

Speaker turn embeddings improve dialogue act classification accuracy.

The approach effectively captures dialogue semantics and speaker dynamics.

Abstract

Dialogue Act (DA) classification is the task of classifying utterances with respect to the function they serve in a dialogue. Existing approaches to DA classification model utterances without incorporating the turn changes among speakers throughout the dialogue, therefore treating it no different than non-interactive written text. In this paper, we propose to integrate the turn changes in conversations among speakers when modeling DAs. Specifically, we learn conversation-invariant speaker turn embeddings to represent the speaker turns in a conversation; the learned speaker turn embeddings are then merged with the utterance embeddings for the downstream task of DA classification. With this simple yet effective mechanism, our model is able to capture the semantics from the dialogue content while accounting for different speaker turns in a conversation. Validation on three benchmark public datasets demonstrates superior performance of our model.