Unsupervised Dialogue Act Induction using Gaussian Mixtures

TL;DR

This paper presents an unsupervised method for dialogue act induction using Gaussian mixture models within a Hidden Markov framework, demonstrating competitive results on a standard corpus without labeled data.

Contribution

It introduces a novel unsupervised approach combining Gaussian mixtures and HMMs for dialogue act induction, outperforming existing unsupervised methods.

Findings

Achieves promising results on Switchboard-DAMSL corpus

Outperforms other unsupervised algorithms

Comparable to supervised baselines

Abstract

This paper introduces a new unsupervised approach for dialogue act induction. Given the sequence of dialogue utterances, the task is to assign them the labels representing their function in the dialogue. Utterances are represented as real-valued vectors encoding their meaning. We model the dialogue as Hidden Markov model with emission probabilities estimated by Gaussian mixtures. We use Gibbs sampling for posterior inference. We present the results on the standard Switchboard-DAMSL corpus. Our algorithm achieves promising results compared with strong supervised baselines and outperforms other unsupervised algorithms.