Modeling Second Language Acquisition with pre-trained neural language models

作者：

Highlights：

• State-of-the-art framework for SLAM that relies on transfer learning.

• Knowledge distillation improves both efficiency and effectiveness.

• The stack-and-finetune approach gives the best results.

摘要

•State-of-the-art framework for SLAM that relies on transfer learning.•Knowledge distillation improves both efficiency and effectiveness.•The stack-and-finetune approach gives the best results.

论文关键词：Second Language Acquisition,Pre-trained Language model,Model Distillation,Fine-tuning,Feature extraction

论文评审过程：Received 16 February 2022, Revised 20 May 2022, Accepted 12 June 2022, Available online 17 June 2022, Version of Record 8 July 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.117871