Unsupervised Learning of Disentangled and Interpretable Representations from Sequential Data

TL;DR

This paper introduces a hierarchical variational autoencoder that learns disentangled, interpretable representations from sequential data, enabling transformations like speaker or content changes and improving speech recognition and verification.

Contribution

It proposes a novel factorized hierarchical model that captures multi-scale information in sequential data without supervision, outperforming baselines in speech tasks.

Findings

Outperforms i-vector baseline in speaker verification

Reduces word error rate by up to 35% in speech recognition

Enables meaningful manipulation of latent variables for speaker and content changes

Abstract

We present a factorized hierarchical variational autoencoder, which learns disentangled and interpretable representations from sequential data without supervision. Specifically, we exploit the multi-scale nature of information in sequential data by formulating it explicitly within a factorized hierarchical graphical model that imposes sequence-dependent priors and sequence-independent priors to different sets of latent variables. The model is evaluated on two speech corpora to demonstrate, qualitatively, its ability to transform speakers or linguistic content by manipulating different sets of latent variables; and quantitatively, its ability to outperform an i-vector baseline for speaker verification and reduce the word error rate by as much as 35% in mismatched train/test scenarios for automatic speech recognition tasks.