Detecting Word Ordering Errors in Chinese Sentences for Learning Chinese as a Foreign Language

Proceedings of COLING 2012: Technical Papers, pages 3003–3018, COLING 2012, Mumbai, December 2012. Detecting Word Ordering Errors in Chinese Sentences for Learning Chinese as a Foreign Language Chi-Hsin Yu, Hsin-Hsi Chen Department of Computer Science and Information Engineering, National Taiwan University #1, Sec.4, Roosevelt Road, Taipei, 10617 Taiwan jsyu@nlg.csie.ntu.edu.tw, hhchen@ntu.edu.tw ABSTRACT Automatic detection of sentence errors is an important NLP task and is valuable to assist foreign language learners. In this paper, we investigate the problem of word ordering errors in Chinese sentences and propose classifiers to detect this type of errors. Word n-gram features in Google Chinese Web 5-gram corpus and ClueWeb09 corpus, and POS features in the Chinese POS- tagged ClueWeb09 corpus are adopted in the classifiers. The experimental results show that integrating syntactic features, web corpus features and perturbation features are useful for word ordering error detection, and the proposed classifier achieves 71.64% accuracy in the experimental datasets. 協助 協助 協助 協助非中文母語學習者偵測中文句子語序錯誤 非中文母語學習者偵測中文句子語序錯誤 非中文母語學習者偵測中文句子語序錯誤 非中文母語學習者偵測中文句子語序錯誤 自動偵測句子錯誤是自然語言處理研究一項重要議題，對於協助外語學習者很有價值。在 這篇論文中，我們研究中文句子語序錯誤的問題，並提出分類器來偵測這種類型的錯誤。 在分類器中我們使用的特徵包括：Google 中文網路 5-gram 語料庫、與 ClueWeb09 語料庫 的中文詞彙 n-grams及中文詞性標注特徵。實驗結果顯示，整合語法特徵、網路語料庫特 徵、及擾動特徵對偵測中文語序錯誤有幫助。在實驗所用的資料集中，合併使用這些特徵 所得的分類器效能可達 71.64%。 KEYWORDS : ClueWeb09, computer-aided language learning, HSK corpus, word ordering error detection KEYWORDS IN L2 : ClueWeb09, 電腦輔助語言學習, HSK 語料庫, 語序錯誤偵測 3003 1 Introduction Non-native language learners usually encounter problems in learning a new foreign language and are prone to generate ungrammatical sentences. NLP systems that detect and correct grammatical errors are important and invaluable to language learners. Error detection, which tells if there exists a special type of errors in a given sentence, is the first step for this kind of applications (Leacock, Chodorow, Gamon, & Tetreault, 2010). Sentences with various types of errors are written by language learners of different backgrounds. The distribution of errors varies across different learner groups. For example, the most frequent error for native English learners is missing comma after introductory element, while the most frequent error for ESL (English as a Second Language) students is comma splice, which joins two sentences by using comma instead of a conjunction (Leacock et al., 2010). In addition, learners may generate sentences with multiple types of errors. This complicates the error detection. To simplify the problem and evaluate the performance of the proposed methods, researchers usually detect one type of error at a time, such as common grammatical errors (Islam & Inkpen, 2011; Wagner, Foster & Genabith, 2007) and prenominal adjective ordering errors (Lin et al., 2009; Malouf, 2000). A training dataset is indispensable for learning an error detection system. In English, there are many well-established corpora for this purpose such as Cambridge Learner Corpus (CLS). In Chinese, Beijing Language and Culture University built an HSK dynamic composition corpus (动态/dynamic 作文/composition 语料库/corpus) and announced a publicly accessible online system to query this corpus1. The HSK corpus contains Chinese compositions written by Chinese language learners in the university. The students’ compositions were collected and annotated with different error types. In this paper, we will deal with the word ordering errors in the HSK corpus. This paper is organized as follows. In Section 2, we survey the related …

Content truncated to 500 words