The article presents research on spoken grammatical error correction (GEC), which is crucial for second language learners. Traditionally, GEC was focused on correcting written text, but with the increasing interest in automating all language learning skills, spoken GEC has become a new area of interest. This study introduces a new end-to-end approach to spoken GEC using a speech recognition model, Whisper, which can replace the entire GEC framework or parts of it. However, the performance of this approach is limited by the lack of available data. The paper also discusses the challenges of providing feedback to candidates using end-to-end systems for spoken GEC.
Publication date: 10 Nov 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2311.05550
