French error type annotation for dictation: A platform with automatic error type annotation for French dictation...

Area of Science:

• Second Language Acquisition
• Computational Linguistics
• Educational Technology

Background:

• Traditional French as a Foreign Language (FFL) dictation exercises offer limited instructional feedback efficiency.
• Automated feedback systems are needed to enhance pedagogical support for FFL learners and educators.

Purpose of the Study:

• To develop and evaluate FRETA-D, an automatic error type annotation platform for FFL dictation.
• To provide intelligent, fine-grained error feedback to FFL learners and teachers.

Main Methods:

• Design-based research approach to build the FRETA-D platform.
• Development of a dataset-independent classifier with 25 error types based on French grammar and common learner errors.
• Validation through expert review by five French teachers and automatic evaluation against human judgments.

Main Results:

• FRETA-D achieved over 85% acceptance rate from expert teachers for error type appropriateness.
• Automatic evaluation showed over 85% consistency with human judgments on 1,009 sentences.
• The platform demonstrated accessibility and utility for 50 Chinese undergraduate FFL learners.

Conclusions:

• FRETA-D offers an effective solution for intelligent pedagogical feedback in FFL dictation.
• The platform facilitates dynamic analysis of learner error patterns.
• Findings reinforce the link between dictation errors and learners' phoneme-grapheme mastery in French.