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

Russian

Текст-к-говору (TTS) обычно оценивается путем транскрипции его выхода с помощью системы автоматического распознавания речи (ASR), поэтому коэффициент ошибки слов также подсчитывает ошибки распознавателя. Мы измеряем эти ошибки на человеческих записях тестовых предложений. На LibriSpeech и FLEURS, распознаватель оценки пропускает 33-67% очень редких слов даже в человеческой речи, и все три распознавателя пропускают 22-42%. Проверка каждого слова с двумя дополнительными семьями распознавателей снижает явную ошибку TTS на слова, которые по крайней мере один распознаватель слышит два- до трех раз (очень редкие слова LibriSpeech: от 10-13% до 3-5%), но очень редкие слова остаются самыми трудными

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

Is this paper real? This page was built from OpenAlex and arXiv records, not generated. Software can verify any reference in one call: GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.31470, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txt

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