Task Adapters

Pre-trained model:

CLC FCE Dataset

The First Certificate in English (FCE) dataset in sequence labeling format, converted by Rei and Yannakoudakis (2016), wher eeach token in a sentence is labelled either as correct or as incorrect. Following the original authors and prior work in error detection, performance on this dataset is measured by an F0.5 score which assigns precision twice as much importance as recall.

Website