Learning Multi-level Dependencies for Robust Word Recognition

TL;DR

This paper presents a robust word recognition framework that captures multi-level dependencies, significantly improving accuracy in noisy conditions by leveraging character-level and word-level neural networks.

Contribution

The paper introduces a novel multi-level dependency framework combining character and word-level neural networks for robust word recognition in noisy environments.

Findings

Outperforms state-of-the-art methods by a large margin

Character-level dependencies are crucial for accurate word recognition

Effective in noisy sentence conditions

Abstract

Robust language processing systems are becoming increasingly important given the recent awareness of dangerous situations where brittle machine learning models can be easily broken with the presence of noises. In this paper, we introduce a robust word recognition framework that captures multi-level sequential dependencies in noised sentences. The proposed framework employs a sequence-to-sequence model over characters of each word, whose output is given to a word-level bi-directional recurrent neural network. We conduct extensive experiments to verify the effectiveness of the framework. The results show that the proposed framework outperforms state-of-the-art methods by a large margin and they also suggest that character-level dependencies can play an important role in word recognition.