Newest papers

Measuring the Verifier's Floor: How Much Apparent Rare-Word Text-to-Speech Error Is ASR Error?

Kirill Borodin, Vasilii Kudryavtsev, Nikita Vasiliev, Grach Mkrtchian · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.31470 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

Text-to-speech (TTS) generalmente se evalúa mediante la transcripción de su salida con un sistema de reconocimiento automático de habla (ASR), por lo que la tasa de error de la palabra también cuenta los errores del reconocedor. Medimos estos errores en las grabaciones humanas de las frases de prueba. En LibriSpeech y FLEURS, el reconocedor de evaluación pierde 33-67% de palabras muy raras incluso en el habla humana, y todos los tres reconocedores pierden 22-42%. Comprobar cada palabra con dos familias de reconocedores más disminuye el error aparente de TTS en palabras que al menos un reconocedor oye de dos a tres veces (muy raras palabras de LibriSpeech: de 10-13% a 3-5%), pero las palabras muy raras siguen siendo las más difíciles. Ignorar las palabras del

Machine translation (Meta M2M100); the English original follows.

Abstract

Text-to-speech (TTS) is usually evaluated by transcribing its output with an automatic speech recognition (ASR) system, so the word error rate also counts recognizer errors. We measure these errors on human recordings of the test sentences. On LibriSpeech and FLEURS, the evaluation recognizer misses 33-67% of very rare words even in human speech, and all three recognizers miss 22-42%. Checking every word with two more recognizer families lowers the apparent TTS error on words that at least one recognizer hears two- to three-fold (very rare LibriSpeech words: from 10-13% to 3-5%), yet very rare words remain the hardest. Ignoring words the recognizer misses in human speech is biased: the other recognizers find TTS errors far more often on these words (4-6% vs. 0.2-0.3%) and overturn about half of the errors this mask keeps. As half of the remaining adjudicated errors may be spelling variants, a listening test is needed.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Speech recognition, Computer science, Word error rate, Word (group theory), Spelling, Active listening

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange