Study Of Correlation Between Objective Factors And Manual Scoring In Chinese Composition Of Overseas Students At Advanced Level

Combined with the current study outcomes of composition scoring in second language acquisition and personal writing teaching practices,we found that the text features,such as vocabulary levels and average sentence of compositions,differ in the compositions at different level.These features can be quantified by statistical methods,so it's called the objective factors.Based on this,we first used the method of correlation to discover whether the data of the objective factors are proportional or inversely proportional to the text scores and the distribution of those factors on texts at different scales,then we built up a scoring model with regression analysis to generate the objective factors,which have significant effects on the manual scoring.Additionally,we tested the reliability of the predicted scores given by the model.Finally,based on the conclusion,we made suggestions to the scoring criteria,scoring methods and second language teaching of writing and textbook compilation.This paper is divided into five chapters:Chapter one: Introduction.This paper first describes the reason of choosing the topic and the object of the study,and then carries on the literature review towards the factors affecting the scores and objectively scoring and summarizes the results,and then explains the purpose and significance of the study.Finally,it clarifies the research methods and the use of data processing methods.Chapter two: The implementation of study on the relationship between objective factors and manual scoring.Before carrying out the experiments,we summarize the characteristics of the selected texts in the predecessors' studies to obtain the frequency of the objective components and integrate their study conclusions,and then 20 objective factors are chosen in this study.After studying the HSK composition scoring system and the scoring forms in the Linguistic Competence Scale in the Common European Framework of Reference for Languages: Learning,Teaching and Evaluation,(Hereinafter referred to as "the framework"),we develop an overall scale for composition evaluation after some revise based on the pre-testing and implement it.In addition,the text sources,scorer information,and sample analysis methods are also described.Chapter three: SPSS processing objective factors and manual score results and analysis;Firstly,we collate the corpus of the study and classify the texts according to the final scores in the samples,and then randomly select one sample from each file for evaluation.We conduct the formal evaluation after the scorers have a certain understanding of the texts.We use the overall scale to get scores,and split the compositions into three scales by the points.Then we calculate the objective factors in each file,and analyze the correlation between these factors and the scores,which is to find out whether the data of these factors is proportional or inversely proportional to the scores.In order to demonstrate the correlation,we utilize the regression method to process the two variables and we obtained the factors that have significant effects on scoring and also a regression equation can be used for automatic scoring practice.Then we compared the scores generated by the equation with those getting from manual scoring to verify the predictive reliability of the equation.Chapter four: Teaching suggestions.In the evaluation,we have obtained objective factors that have significant influence on the final scores.It is considerable that these factors should be included in the scoring standard.In addition,we've analyzed the raters' severity for the texts on different grades and the degree of consistency between the rating stuff,and also considering the factors that significantly affect the scores;we make suggestions for the scoring method.Chapter five: The fifth chapter: Conclusions.The conclusions are divided into two parts.Firstly,the relationship between objective factors and scores.Specifically,The higher the score,the higher the proportion of words,the less the wrong character,the wrong word and the wrong sentence,the more the total number of sentences and clauses,the longer the average sentences and clauses.Secondly,the factors,which significantly impact the scoring,are generated.In the rank low is the difficulty of the text vocabulary,in mid-range is the difficulty of the lexical and article fluency,and in high-grade compositions is the difficulty of the word,the length of the article,the article correctness.In the follow-up study,we think that the study samples should be increased and rating experts can be introduced to improve the experiment.