Compositional Sequence Labeling Models for Error Detection in Learner Writing

TL;DR

This paper introduces neural network-based compositional models for error detection in learner writing, demonstrating their effectiveness through experiments and integration into a self-assessment system.

Contribution

It is the first to systematically compare neural architectures for learner error detection and proposes a bidirectional LSTM framework that outperforms previous methods.

Findings

Outperforms other participants on CoNLL-14 dataset

Achieves error detection performance comparable to human annotators

Successfully integrated into a real-world self-assessment system

Abstract

In this paper, we present the first experiments using neural network models for the task of error detection in learner writing. We perform a systematic comparison of alternative compositional architectures and propose a framework for error detection based on bidirectional LSTMs. Experiments on the CoNLL-14 shared task dataset show the model is able to outperform other participants on detecting errors in learner writing. Finally, the model is integrated with a publicly deployed self-assessment system, leading to performance comparable to human annotators.