ICC 25427
INSTITUTO DE INVESTIGACION EN CIENCIAS DE LA COMPUTACION
Unidad Ejecutora - UE
congresos y reuniones científicas
Título:
EpaDB: A Database for Development of Pronunciation Assessment Systems
Autor/es:
LEONARDO BRAMBILLA; LUCIANA FERRER; JAZMIN VIDAL
Lugar:
Graz
Reunión:
Congreso; Interspeech 2019; 2019
Institución organizadora:
ISCA
Resumen:
In this paper, we describe the methodology for collecting andannotating a new database designed for conducting research anddevelopment on pronunciation assessment. While a significantamount of research has been done in the area of pronunciationassessment, to our knowledge, no database is available for pub-lic use for research in the field. Considering this need, we cre-ated EpaDB (English Pronunciation by Argentinians Database),which is composed of English phrases read by native Span-ish speakers with different levels of English proficiency. Therecordings are annotated with ratings of pronunciation qualityat phrase-level and detailed phonetic alignments and transcrip-tions indicating which phones were actually pronounced by thespeakers. We present inter-rater agreement, the effect of eachphone on overall perceived non-nativeness, and the frequencyof specific pronunciation errors.
