Paper: Generalized Character-Level Spelling Error Correction

ACL ID P14-2027
Title Generalized Character-Level Spelling Error Correction
Venue Annual Meeting of the Association of Computational Linguistics
Session Main Conference
Year 2014

We present a generalized discrimina- tive model for spelling error correction which targets character-level transforma- tions. While operating at the charac- ter level, the model makes use of word- level and contextual information. In con- trast to previous work, the proposed ap- proach learns to correct a variety of er- ror types without guidance of manually- selected constraints or language-specific features. We apply the model to cor- rect errors in Egyptian Arabic dialect text, achieving 65% reduction in word error rate over the input baseline, and improv- ing over the earlier state-of-the-art system.