高级阶段留学生汉语书面表达中客观性因素和人工评分的相关性研究

发布时间：2018-04-17 21:29

本文选题：高级阶段留学生 + 汉语书面表达　；参考：《上海外国语大学》2017年硕士论文

【摘要】：结合目前第二语言学习领域作文评判的成果和个人写作教学实践经验,我们发现:处于不同分数段的作文,在词汇难度、平均句长等浅层文本特征上存在差异。这些特征可用量化的方法进行统计,本文称其为客观性因素。本文以此着手,首先运用相关分析的方法来探讨这些客观性因素和其所在文本分数是否成正比或反比,以及客观性因素在不同等级作文上的分布差异;之后通过回归分析建立的评分模型获得对人工评分有显著影响的客观性因素并检验该模型预测分数的信度;最后基于结论对评分标准、评分方法、写作教学及教材编写提出了建议。本论文共分五章:第一章绪论;本文首先叙述了论文的选题缘由和研究对象,然后进行了针对作文评分影响因素和客观化评分的文献回顾并对其进行了总结,之后说明了研究的目的和意义,最后阐明本文研究思路和运用的数据处理方法。第二章客观性因素和人工评分关系研究实施;在实验之前,我们通过汇总前人研究中选取的文本特征,获得对各客观成分的研究频率,并结合前人的研究结论,共选取了20个客观性因素;在学习了HSK作文评分标准和《欧洲语言共同参考框架:学习、教学、评估》(以下简称《框架》)中的语言能力量表等之后,我们制定了用于本次研究的整体量表,并在试评后作出更改,最后用于作文测评。此外,我们对文本来源、评分员信息及样本分析方法也进行了说明。第三章SPSS处理客观性因素和人工评分结果及分析;首先我们整理了研究语料,依据样本分数对文本进行了分档,然后在各档中随机抽取一篇用于试评,在评分员对样本语言质量有一定认识后开始正式测评。评分时我们选用整体量表获得作文总分,并依据得分将作文分成三档,随后我们统计了各档中的客观性因素,并分析了这些因素的数据是否与得分成正比或反比,即两者是否具有相关关系。为了更详细地说明客观性因素与总分的关系,我们将样本中的客观性因素和对应总分进行回归分析,获得对总分有有效解释力的显著因素,并随机选取样本人工再测,将所得分数与用回归方程得到的分数进行T检验以验证回归方程的预测信度。第四章评分建议;在测评中,我们获得了对总分有显著影响的客观性因素,我们认为这些因素应纳入评分标准中;此外,我们也分析了评分员对不同档次文本的严厉度和评分员之间的一致性程度,同时考虑了显著影响评分员的因素,对评分方法提出了建议。第五章结语;本次研究结论分为两部分。一是得出客观性因素与总分之间的关系,具体为作文分数越高,非常用词比重越高,错字数、错词数和错句数越少,总句数、分句数越多,平均分句长、平均句长等越长。二是得到对评分有显著影响的因素。在低档是词汇难度,中档是词汇难度和文章流利度,高档是词汇难度、文章长度和文章正确度。在后续的研究中,我们认为可以增加研究样本,采用评分专家进行评分,以改善实验。
[Abstract]:Combining the second language learning achievement and personal evaluation of composition writing teaching practice experience, we find that in different levels of composition in lexical difficulty, average sentence length and other characteristics of the text of shallow differences. These features can be quantified by statistics, the paper called for objectivity factors. This method to we use correlation analysis to explore these factors and objectivity of the text is proportional to the fraction or inverse, distribution in different levels on the composition and objective factors; the objective factors have significant influence on artificial scoring score model regression analysis to establish the prediction model and test the reliability of scores; finally, based on the conclusion the scoring method of standard for evaluation, writing, teaching and textbook compiling suggestions. This paper is divided into five chapters: the first chapter is the introduction firstly; Describes the researchbackground and object, then according to the grading factors and objective scoring of the literature was reviewed and summarized them, then explains the purpose and significance of the research, finally expounds the idea and method of data processing in this paper. The second chapter study on the relationship between artificial factors and objective scores; before the experiment, we through the summary of text feature selection in previous studies, research on the frequency of each objective component, combined with the previous research conclusions, we select 20 objective factors; in the study of the composition of HSK standard for evaluation and "common European Framework of reference for languages: learning, teaching, assessment (hereinafter referred to as < > after the framework >) language ability scale, we developed for the overall volume of the study table, and make changes in the assessment test, finally used for writing assessment. In addition, we in this paper The source of information, rater and sample analysis method is also described. The third chapter SPSS objective factors and artificial scoring results and analysis; first we analyzed the research data, according to the sample scores were classified to the text in the file, then randomly selected for a trial evaluation, began formal evaluation in the score who have some knowledge of quality. After the sample language score when we choose the whole scale to obtain composition scores, and scores according to the composition into third gear, then we counted each file in the objective factors, and analyzed the data for these factors are directly proportional or inversely with scores, which both are related to more. A detailed description of the relationship between the objective factor and the total score, we will be in a sample of objective factors and the corresponding score regression analysis, get effective explanation to the total score of the significant factors, and with the The sample machine manual test, the scores and scores of the regression equation to predict the reliability of T test of the regression equation. The fourth chapter score recommendations; in the evaluation, we obtain objective factors have significant influence on the total score, we believe that these factors should be included in the standard for evaluation; in addition, we also analyze between raters for different grades of the text of the severity and rater consistency at the same time, considering the significant influence factors of raters, suggestions on scoring method is proposed. The fifth chapter is the conclusion; the conclusion of the study is divided into two parts. One is the relationship between the objective factors and the total score, the writing scores for specific the higher, the higher the proportion of words is very wrong, wrong words, words and sentences are few, the total number of sentences, clauses number, average clause length, average sentence length and longer. Two is obtained had a significant effect on the score for Low grade is lexical difficulty. Middle grade is lexical difficulty and article fluency. High-end is lexical difficulty, article length and article accuracy. In subsequent research, we think that we can increase research samples and score experts to score, so as to improve the experiment.

【学位授予单位】：上海外国语大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：H195

【参考文献】