Using Transfer Learning to Automatically Mark L2 Writing Texts
The use of transfer learning in Natural Language Processing (NLP) has grown over the last few years. Large, pre-trained neural networks based on the Transformer architecture are one example of this, achieving state-of-theart performance on several commonly used performance benchmarks, often when finetuned on a downstream task. Another form of transfer learning, Multitask Learning, has also been shown to improve performance in Natural Language Processing tasks and increase model robustness.