Spelling Error Correction Using a Nested RNN Model and Pseudo Training Data

TL;DR

This paper introduces a nested RNN model for English spelling error correction that leverages pseudo data for training, achieving superior performance without traditional feature engineering or noisy channel models.

Contribution

The paper presents a novel nested RNN architecture combined with pseudo data generation for effective spelling error correction, bypassing traditional noisy channel approaches.

Findings

Outperforms existing spelling correction systems

Effective use of pseudo data improves accuracy

End-to-end training simplifies the correction process

Abstract

We propose a nested recurrent neural network (nested RNN) model for English spelling error correction and generate pseudo data based on phonetic similarity to train it. The model fuses orthographic information and context as a whole and is trained in an end-to-end fashion. This avoids feature engineering and does not rely on a noisy channel model as in traditional methods. Experiments show that the proposed method is superior to existing systems in correcting spelling errors.