The Analysis Of The Text Quality Of Chinese Heritage Language Composition And Its Relation To Comprehensive Scoring

This study takes the figure compositions as samples, which are from the time-limited task of writing a Chinese composition on the same title by Grade 2, 4 and 6 students in primary school of China, Malaysian and Indonesia. It assesses the text quality of composition samples through quantitative analysis and comprehensive scoring, and analyzes the relation between text linguistic quality and the comprehensive scores of these samples. The conclusions are as follow.(1) All the compositions’ text quality and comprehensive scores are improving as their grades are increasing. The quantity and difficulty of Chinese characters, vocabularies, and the accuracy rate of sentences are significantly improving; the difficulty rate of sentences is slightly increasing. The error rate of Chinese characters and words are reducing. The overall scores are increasing.(2) Comparing the samples from different learners from three countries, the Malaysia Chinese heritage language learner samples’ quality is a little higher than Chinese native learners’, however, the Indonesia samples’ quality is much less than the other two countries’. The difficulty of Chinese characters, vocabularies, the difficulty and accuracy rate of sentences, and the overall scores of Malaysia samples are all higher than native samples. Nevertheless the error rate of Chinese characters and vocabularies is higher than native samples. The quantity and difficulty of Chinese characters, vocabularies, the difficulty and accuracy rate of sentences of Indonesia samples are far less than the other two samples. Their error rate of Chinese characters and vocabularies is much higher than the other two; and overall scores are far behind others, too.(3)The text quality of composition can predict its score to some extent, which could be refered by automatic scoring of Chinese composition. We find that all the 52 variables extracted from text are related to scores by correlation analysis. And we choose 4 optimal variables to form an equation through stepwise multiple regression analysis: score=46.025+0.241×the amount of no-repeat Chinese characters in Jiaji-43.371×the error rate of no-repeat Chinese characters. This equation can explain 62.4% of the composition scores.