Abstract This study reports classroom research on corpus use in second-language (L2) error correction to explore how information from corpus data influences L2 writing. As the error types for which corpus consultation is helpful remain uncertain, this study examined the error types for which data-driven learning (DDL) provides good error correction in L2 writing to maximize the benefits of corpus use in L2 class instruction. First, 55 undergraduate students from a university in Tokyo wrote an essay in 25 min without access to reference resources and received teacher or peer feedback on their errors. They then performed revision tasks for 15 min with either use or non-use of reference resources (corpus or dictionary). This procedure was carried out 9–11 times during the term. Error analysis found that error types affected how frequently and accurately learners corrected their errors using corpus data. Corpus use allowed easy access to the exact target phrases and frequency information of co-occurring words, which especially helped participants correct omission errors, and corpus consultation to correct omission errors increased over time. Teachers should consider error types so that DDL can promote accurate error correction in L2 writing and serve as a practical option in L2 classrooms.

中文翻译：

---

使用语料库时错误类型如何影响 L2 纠错的准确性

摘要 本研究报告了在二语 (L2) 纠错中使用语料库的课堂研究，以探讨语料库数据中的信息如何影响 L2 写作。由于语料库咨询有帮助的错误类型仍然不确定，本研究检查了数据驱动学习 (DDL) 在 L2 写作中提供良好纠错的错误类型，以最大限度地提高语料库在 L2 课堂教学中使用的好处。首先，东京一所大学的 55 名本科生在 25 分钟内写了一篇论文，没有参考资源，并收到了老师或同行对他们错误的反馈。然后，他们在使用或不使用参考资源（语料库或字典）的情况下执行 15 分钟的修订任务。该程序在学期内进行了 9-11 次。错误分析发现，错误类型会影响学习者使用语料库数据纠正错误的频率和准确度。使用语料库可以轻松获取准确的目标短语和共现词的频率信息，这尤其有助于参与者纠正遗漏错误，并且随着时间的推移，纠正遗漏错误的语料库咨询会增加。教师应考虑错误类型，以便 DDL 可以促进 L2 写作中的准确纠错，并作为 L2 课堂的实用选择。