Domain-shift Conditioning using Adaptable Filtering via Hierarchical Embeddings for Robust Chinese Spell Check

TL;DR

This paper introduces a scalable, adaptable filtering method using hierarchical embeddings for Chinese spell check, effectively handling domain shifts and reducing the need for handcrafted confusion sets, achieving state-of-the-art results.

Contribution

The paper presents a novel hierarchical embedding-based filtering approach that automatically adapts to new error domains in Chinese spell check, overcoming limitations of fixed confusion sets.

Findings

Outperforms competitive baselines on Chinese Spelling Check datasets

Achieves state-of-the-art results on 2014 and 2015 Bake-off datasets

Effectively handles domain shifts and error sparsity

Abstract

Spell check is a useful application which processes noisy human-generated text. Spell check for Chinese poses unresolved problems due to the large number of characters, the sparse distribution of errors, and the dearth of resources with sufficient coverage of heterogeneous and shifting error domains. For Chinese spell check, filtering using confusion sets narrows the search space and makes finding corrections easier. However, most, if not all, confusion sets used to date are fixed and thus do not include new, shifting error domains. We propose a scalable adaptable filter that exploits hierarchical character embeddings to (1) obviate the need to handcraft confusion sets, and (2) resolve sparsity problems related to infrequent errors. Our approach compares favorably with competitive baselines and obtains SOTA results on the 2014 and 2015 Chinese Spelling Check Bake-off datasets.