FAB: The French Absolute Beginner Corpus for Pronunciation Training

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 6613–6620 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 6613 FAB: The French Absolute Beginner Corpus for Pronunciation Training Sean Robertson1, Cosmin Munteanu2, Gerald Penn1 1Department of Computer Science, University of Toronto 40 St. George St., Toronto, ON, Canada 2Institute of Communication, Culture, Information and Technology, University of Toronto Mississauga 3359 Mississauga Rd., Mississauga, ON, Canada sdrobert@cs.toronto.edu, cosmin.munteanu@utoronto.ca, gpenn@cs.toronto.edu Abstract We introduce the French Absolute Beginner (FAB) speech corpus. The corpus is intended for the development and study of Computer-Assisted Pronunciation Training (CAPT) tools for absolute beginner learners. Data were recorded during two experiments focusing on using a CAPT system in paired role-play tasks. The setting grants FAB three distinguishing features from other non-native corpora: the experimental setting is ecologically valid, closing the gap between training and deployment; it features a label set based on teacher feedback, allowing for context-sensitive CAPT; and data have been primarily collected from absolute beginners, a group often ignored. Participants did not read prompts, but instead recalled and modified dialogues that were modelled in videos. Unable to distinguish modelled words solely from viewing videos, speakers often uttered unintelligible or out-of-L2 words. The corpus is split into three partitions: one from an experiment with minimal feedback; another with explicit, word-level feedback; and a third with supplementary read-and-record data. A subset of words in the first partition has been labelled as more or less native, with inter-annotator agreement reported. In the explicit feedback partition, labels are derived from the experiment’s online feedback. The FAB corpus is scheduled to be made freely available by the end of 2020. Keywords: L2 Corpus, Non-Native, French, Beginner, Ecological Validity, Pronunciation Training, CAPT, CALL, ASR 1. Introduction Computer-Assisted Language Learning has the potential to greatly ease the burden of educators. There has been consid- erable effort put into applying speech processing technolo- gies to Computer-Assisted Pronunciation Training (CAPT) (Eskenazi, 2009), of which speech recognition is an espe- cially important component. There are a number of difficult challenges involved in building these systems — chiefly the need for learner data collected in an appropriate manner. In this paper, we introduce the French Absolute Beginner (FAB) corpus. Speech data from the FAB corpus were collected over the course of two experiments (Robertson et al., 2016; Robertson et al., 2018). These experiments were designed around co-operative role-playing tasks similar to those found in language-learning classrooms. The results of the experiment in Robertson et al. (2018) highlighted the importance of careful consideration and adherence to the environments in which CAPT systems are to be deployed. In the same spirit, FAB has three distinguishing factors that motivate its use in the development of CAPT systems: • FAB was collected over the course of two ecologi- cally valid experiments, which more closely resemble a classroom CAPT intervention. • One partition of data is labelled according to the native vs. non-native paradigm; another is labelled according to teacher feedback. • FAB …

Content truncated to 500 words