Grammatical Error Correction with Alternating Structure Optimization

Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics, pages 915–923, Portland, Oregon, June 19-24, 2011. c⃝2011 Association for Computational Linguistics Grammatical Error Correction with Alternating Structure Optimization Daniel Dahlmeier1 and Hwee Tou Ng1,2 1NUS Graduate School for Integrative Sciences and Engineering 2Department of Computer Science, National University of Singapore {danielhe,nght}@comp.nus.edu.sg Abstract We present a novel approach to grammatical error correction based on Alternating Struc- ture Optimization. As part of our work, we introduce the NUS Corpus of Learner En- glish (NUCLE), a fully annotated one mil- lion words corpus of learner English available for research purposes. We conduct an exten- sive evaluation for article and preposition er- rors using various feature sets. Our exper- iments show that our approach outperforms two baselines trained on non-learner text and learner text, respectively. Our approach also outperforms two commercial grammar check- ing software packages. 1 Introduction Grammatical error correction (GEC) has been rec- ognized as an interesting as well as commercially attractive problem in natural language process- ing (NLP), in particular for learners of English as a foreign or second language (EFL/ESL). Despite the growing interest, research has been hindered by the lack of a large annotated corpus of learner text that is available for research purposes. As a result, the standard approach to GEC has been to train an off-the-shelf classifier to re-predict words in non-learner text. Learning GEC models directly from annotated learner corpora is not well explored, as are methods that combine learner and non-learner text. Furthermore, the evaluation of GEC has been problematic. Previous work has either evaluated on artificial test instances as a substitute for real learner errors or on proprietary data that is not available to other researchers. As a consequence, existing meth- ods have not been compared on the same test set, leaving it unclear where the current state of the art really is. In this work, we aim to overcome both problems. First, we present a novel approach to GEC based on Alternating Structure Optimization (ASO) (Ando and Zhang, 2005). Our approach is able to train models on annotated learner corpora while still tak- ing advantage of large non-learner corpora. Sec- ond, we introduce the NUS Corpus of Learner En- glish (NUCLE), a fully annotated one million words corpus of learner English available for research pur- poses. We conduct an extensive evaluation for ar- ticle and preposition errors using six different fea- ture sets proposed in previous work. We com- pare our proposed ASO method with two baselines trained on non-learner text and learner text, respec- tively. To the best of our knowledge, this is the first extensive comparison of different feature sets on real learner text which is another contribution of our work. Our experiments show that our pro- posed ASO algorithm significantly improves over both baselines. It also outperforms two commercial grammar checking software packages in a manual evaluation. The remainder of this paper is organized as fol- lows. The next section reviews related work. Sec- tion 3 describes …

Content truncated to 500 words