Anomaly-Based Annotation Errors Detection in TTS Corpora

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F49777513%3A23520%2F15%3A43926613" target="_blank" >RIV/49777513:23520/15:43926613 - isvavai.cz</a>

Výsledek na webu

<a href="http://www.isca-speech.org/archive/interspeech_2015/i15_0314.html" target="_blank" >http://www.isca-speech.org/archive/interspeech_2015/i15_0314.html</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Anomaly-Based Annotation Errors Detection in TTS Corpora

Popis výsledku v původním jazyce

In this paper we adopt several anomaly detection methods to detect annotation errors in single-speaker read-speech corpora used for text-to-speech (TTS) synthesis. Correctly annotated words are considered as normal examples on which the detec- tion methods are trained. Misannotated words are then taken as anomalous examples which do not conform to normal patterns of the trained detection models. Word-level feature sets including basic features derived from forced alignment, and various acoustic, spectral, phonetic, and positional features were examined. Dimensionality reduction techniques were also applied to reduce the number of features. The first results with F1 score being almost 89% show that anomaly detection could help in detecting annotation errors in read-speech corpora for TTS synthesis.

Název v anglickém jazyce

Anomaly-Based Annotation Errors Detection in TTS Corpora

Popis výsledku anglicky

In this paper we adopt several anomaly detection methods to detect annotation errors in single-speaker read-speech corpora used for text-to-speech (TTS) synthesis. Correctly annotated words are considered as normal examples on which the detec- tion methods are trained. Misannotated words are then taken as anomalous examples which do not conform to normal patterns of the trained detection models. Word-level feature sets including basic features derived from forced alignment, and various acoustic, spectral, phonetic, and positional features were examined. Dimensionality reduction techniques were also applied to reduce the number of features. The first results with F1 score being almost 89% show that anomaly detection could help in detecting annotation errors in read-speech corpora for TTS synthesis.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

20205 - Automation and control systems

Projekt

<a href="/cs/project/TA01030476" target="_blank" >TA01030476: Inteligentní technologie pro zvýšení bezpečnosti letového provozu</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2015

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Proceedings of the 16th Annual Conference of the International Speech Communication Association (Interspeech 2015)

ISBN

978-1-5108-1790-6

ISSN

2308-457X

e-ISSN

—

Počet stran výsledku

5

Strana od-do

314-318

Název nakladatele

Curran Associates, Inc.

Místo vydání

New York

Místo konání akce

Dresden, Germany

Datum konání akce

6. 9. 2015

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000380581600064