Predicting Multidimensional Subjective Ratings of Children' Readings from the Speech Signals for the Automatic Assessment of Fluency

Gérard Bailly (1) , Erika Godde (1, 2) , Anne-Laure Piat-Marchand (1) , Marie-Line Bosse (2)

1 GIPSA-CRISSP - GIPSA - Cognitive Robotics, Interactive Systems, & Speech Processing
2 LPNC - Laboratoire de Psychologie et NeuroCognition

The objective of this research is to estimate multidimensional subjective ratings of the reading performance of young readers from signalbased objective measures. We here combine linguistic features (number of correct words, repetitions, deletions, insertions uttered per minute. . .) with phonetic features. Expressivity is particularly difficult to predict since there is no unique golden standard. We here propose a novel framework for performing such an estimation that exploits multiple references performed by adults and demonstrate its efficiency using recordings of 273 pupils.

Mots clés: reading aloud; text-to-speech alignment; prosody; children voices

Domaines: Sciences cognitives

Gérard Bailly, Erika Godde, Anne-Laure Piat-Marchand, Marie-Line Bosse. Predicting Multidimensional Subjective Ratings of Children' Readings from the Speech Signals for the Automatic Assessment of Fluency. LREC 2020 - 12th Conference on Language Resources and Evaluation (LREC 2020), May 2020, Marseille, France. pp.317-322. ⟨hal-03039160⟩