Skip to Main content Skip to Navigation
Conference papers

Mesures automatiques de parole non-native : exploration pilote d'un corpus d'apprenants japonais de français et différenciation de niveaux

Abstract : This study focuses on the automatic evaluation of the speech of Japanese learners of French. Our data are semi-spontaneous oral productions of learners from the CLIJAF corpus. We evaluate different linguistic levels (pronunciation, lexicon, syntax, discourse) using automatic assessment tools. Our first objective is to study whether these measurements are in line with the general level of proficiency of the learners, as defined by the CEFRL (A2 to C2). Our second objective is to explore the possibility of predicting this level using the most relevant linguistic parameters, with an unsupervised machine learning algorithm. The results confirm the relevance of our objectives and show that automatically extracted parameters such as lexical diversity, the average length of speech turns (syntactic level), the proportion of connectives in the discourse and several parameters related to fluency can contribute to the prediction of learners’ level.
Complete list of metadata

https://hal.archives-ouvertes.fr/hal-03700378
Contributor : Verdiana De Fino Connect in order to contact the contributor
Submitted on : Tuesday, June 21, 2022 - 10:27:59 AM
Last modification on : Tuesday, September 13, 2022 - 6:13:23 PM
Long-term archiving on: : Thursday, September 22, 2022 - 6:57:37 PM

File

Mesures automatiques de parole...
Files produced by the author(s)

Identifiers

  • HAL Id : hal-03700378, version 1

Citation

Verdiana De Fino, Lionel Fontan, Julien Pinquier, Corentin Barcat, Isabelle Ferrané, et al.. Mesures automatiques de parole non-native : exploration pilote d'un corpus d'apprenants japonais de français et différenciation de niveaux. 34émes Journées d'Études sur la Parole (JEP2022), Jun 2022, Noirmoutier, France. pp.1-10. ⟨hal-03700378⟩

Share

Metrics

Record views

40

Files downloads

10