本文通过应用语言学与语言教学的视角，采用实验教学的方法，研究了不同作文评改模式对于大学生英语写作水平的影响。通过写作能力、情感因素、学习成绩三个方面因素的控制，本项研究旨在揭示不同评改方式对于大学英语写作教学的反拨作用。具体来说，本项研究通过“团队教学+过程写作”的方法，采用手动、半自动、自动三种评改系统，对于普快、中快、高快三种语言水平的大学生群体，进行了为期17周的写作实验教学。从论文结构来看，本文首先介绍了国内外不同评改模式在英语写作教学中应用情况，其次展示了手动、半自动、自动评改模式的具体内容与操作方法，再次，阐述了本次实验教学的总体设计、理论框架、教学方法、实验过程、受试主体、测试方式、数据分类。本项研究的实验数据主要来自于两个方面：1）学生受试英语写作水平的前测与后测成绩；2）实验教学问卷调查结果分析。通过这两大类数据的整理分析，最终获得原始数据3.5万条。基于这些实验数据，本文首先从整体实验效果对比、评改类型分组对比、语言水平分组对比这三个方面，对学生的写作能力进步进行了分析，找出了组内和组间的差异，同时结合受试的英语考试成绩和实验教学问卷调查结果，进行了同情感因素与学习体验的相关性分析。统计分析结果显示，合理应用自动评改模式，相对于手动评改模式而言，不仅可以快速地提高大学生的写作水平，而且对于其语言表达的准确性和流畅性方面也有提升。其次，问卷调查结果表明，大学英语写作教学中，引入机器自动评改与教师人工评议相结合的方法，可以更好地改进学生的学习体验，舒缓其写作焦虑感。再次，不同评改模式形成的能力差异，具有可迁移性的特点。这一点，在受试期末大学英语课程考试中得到了印证——写作课上进步大的同学，大学英语课期末写作考试部分的得分也高。最后，本项研究证明，技术不是万能的，纯粹的自动化评改并不可取，生态化的大学英语写作教学改革，需要硬件（教学条件）、软件（教学系统）、人件（教师投入）三个方面的有机结合，才能获得更好的教学效果。
We took a perspective from applied linguistics and language teaching to look at the efficiency of different feedback devices in the teaching of English writing. Three major factors (writing ability, affective impact, and academic records) were used to measure the influence of feedback on college English writing. More specifically, we adopted a "team teaching + process writing" approach which was technically assisted by manual, semi-automated, and automated feedback devices to test the subjects' improvement. The experiment lasted 17 weeks and involved general, intermediate, and advanced levels of students. We first discussed in this thesis the use of feedback in the teaching of English writing. A brief introduction of the three feedback patterns was then given. The information related to general design, theoretical framework, methodology, experimental procedures, subjects, testing methods, and data classification was provided as well. The scores obtained by the subjects on their pretest and posttest were used to measure their improvement, while the results of a questionnaire survey were used to show the affective and experiential impacts on their writing improvement. 35 000 items of raw data were obtained and then categorized in this experiment. Based on these data, we discussed the general improvement of the student writers, compared the effect of separate feedback devices, and analyzed the influence of language levels, with the within-group difference and the between-group difference being examined. Furthermore, based on the results of the questionnaire survey and the academic records, correlation tests were conducted to look at the relationship of affective and experiential factors with the improvement of the student writers. The statistical results indicate that the automated feedback worked better than the manual feedback: It helped improve student writers' performance and enhance accuracy and fluency of their writing. Our other findings include: A combination of automated feedback and instructor review was a good way to enhance the sense of learners' happiness by reducing their anxiety; their improvement was transferable as another English test revealed. We tentatively conclude that an ecological use of hardware (teaching facilities), software (computerized devices), and "human ware" (instructors' involvement) jointly produced an optimal teaching results.
