Application of text clustering in automatic scoring of College English composition

摘要

Writing is an important way to measure the examinees' language knowledge and word organization ability in large-scale language examination. However, the method based on manual scoring has the disadvantages of consuming huge manpower, material and financial resources, strong subjectivity and large error in scoring. In this paper, through the algorithm modeling, clear clustering steps, feature vector extraction, clustering calculation, empirical analysis and other steps, the analysis shows that in the case of College Students' English composition is not completely off topic, text clustering algorithm can well identify the degree of composition in line with the meaning of the topic, and through the establishment of threshold value, teachers can set the judgment authority of the algorithm, which can solve the problem of manual work to improve the accuracy of College English automatic scoring.