A purpose-built online error detection tool was developed to provide genre-specific corpus-based feedback on errors occurring in draft research articles and graduation theses. The primary envisaged users were computer science majors studying at a public university in Japan. This article discusses the development and evaluation of this interactive, multimodal tool. An in-house learner corpus of graduation theses was annotated for errors that affect the accuracy, brevity, clarity, objectivity and formality of scientific research writing. Software was developed to identify the errors discovered and provide learners with actionable advice and multimodal explanations in both English and Japanese. Qualitative evaluation received in usability studies and focus groups from both teachers and students was extremely positive. Preliminary quantitative evaluation of the effectiveness of the error detector was conducted. Through this pedagogic tool, learners can receive immediate actionable feedback on potential errors, and their teachers no longer feel obliged to check for common genre-specific errors.

中文翻译：

---

使用多模态反馈进行特定类型的错误检测

开发了一种专门构建的在线错误检测工具，以针对研究文章草稿和毕业论文中发生的错误提供基于特定类型语料库的反馈。主要设想的用户是在日本公立大学学习的计算机科学专业的学生。本文讨论了这种交互式多模式工具的开发和评估。毕业论文的内部学习者语料库针对影响科学研究写作的准确性、简洁性、清晰度、客观性和形式性的错误进行了注释。开发了软件以识别发现的错误，并以英语和日语为学习者提供可操作的建议和多模式解释。在可用性研究和焦点小组中收到的来自教师和学生的定性评价非常积极。对错误检测器的有效性进行了初步定量评估。通过这种教学工具，学习者可以立即收到关于潜在错误的可操作反馈，他们的老师不再觉得有义务检查常见的特定类型错误。