Predicting pairwise preferences between TTS audio stimuli using parallel ratings data and anti-symmetric twin neural networks

TL;DR

This paper introduces a neural network model that predicts listener preferences between two speech stimuli, leveraging pairwise ratings data and anti-symmetric twin neural networks to improve subjective preference prediction accuracy.

Contribution

The paper presents a novel anti-symmetric twin neural network approach for predicting pairwise preferences in speech stimuli, utilizing converted rating data from MUSHRA tests.

Findings

Model outperforms existing MOS prediction models in preference tasks.

Effective handling of unaligned stimuli with attention and recurrent neural nets.

Robust performance across diverse TTS systems and speaker data.

Abstract

Automatically predicting the outcome of subjective listening tests is a challenging task. Ratings may vary from person to person even if preferences are consistent across listeners. While previous work has focused on predicting listeners' ratings (mean opinion scores) of individual stimuli, we focus on the simpler task of predicting subjective preference given two speech stimuli for the same text. We propose a model based on anti-symmetric twin neural networks, trained on pairs of waveforms and their corresponding preference scores. We explore both attention and recurrent neural nets to account for the fact that stimuli in a pair are not time aligned. To obtain a large training set we convert listeners' ratings from MUSHRA tests to values that reflect how often one stimulus in the pair was rated higher than the other. Specifically, we evaluate performance on data obtained from twelve MUSHRA evaluations conducted over five years, containing different TTS systems, built from data of different speakers. Our results compare favourably to a state-of-the-art model trained to predict MOS scores.