Evaluating Long-form Text-to-Speech: Comparing the Ratings of Sentences and Paragraphs

TL;DR

This paper examines three different evaluation methods for long-form text-to-speech systems, revealing that traditional sentence-level evaluation is insufficient and multiple methods are necessary for accurate assessment.

Contribution

It introduces and compares three evaluation approaches for long-form speech synthesis, highlighting their differences and the need for multiple assessments.

Findings

Evaluation outcomes differ across methods

Results do not always correlate between methods

Multiple evaluation approaches are necessary for accurate assessment

Abstract

Text-to-speech systems are typically evaluated on single sentences. When long-form content, such as data consisting of full paragraphs or dialogues is considered, evaluating sentences in isolation is not always appropriate as the context in which the sentences are synthesized is missing. In this paper, we investigate three different ways of evaluating the naturalness of long-form text-to-speech synthesis. We compare the results obtained from evaluating sentences in isolation, evaluating whole paragraphs of speech, and presenting a selection of speech or text as context and evaluating the subsequent speech. We find that, even though these three evaluations are based upon the same material, the outcomes differ per setting, and moreover that these outcomes do not necessarily correlate with each other. We show that our findings are consistent between a single speaker setting of read paragraphs and a two-speaker dialogue scenario. We conclude that to evaluate the quality of long-form speech, the traditional way of evaluating sentences in isolation does not suffice, and that multiple evaluations are required.