计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (9): 151-155.DOI: 10.3778/j.issn.1002-8331.1611-0502

• 模式识别与人工智能 • 上一篇    下一篇

集中趋势自适应增强的英语作文评分算法

李  婷,张景祥   

  1. 济南大学 信息科学与工程学院,济南 250022
  • 出版日期:2018-05-01 发布日期:2018-05-15

Adaptive boosting with central tendency algorithm for English essay scoring

LI Ting, ZHANG Jingxiang   

  1. School of Information Science and Engineering, University of Jinan, Jinan 250022, China
  • Online:2018-05-01 Published:2018-05-15

摘要: 为更好地对英语作文进行智能评分,提出了一种改进算法Adaboost/CT。算法以机器筛选得到的主成分作为弱分类器集,通过集中趋势的方法改进了自适应增强技术。这样既避免了过拟合问题,也解决了弱分类器叠加错误陷阱。实验表明该算法能有效地应用于英语作文智能评分系统,且与人工评分相比,其邻接准确率为94%,误差均小于20%且不存在奇异值性误差。该算法在智能评分和机器学习方面具有理论和实用价值。

关键词: 智能作文评分, Adaboost/CT, 指标体系, 分层分析

Abstract: An improved algorithm of Adaboost/CT is proposed to find a better way in intelligent English essay assessment. It uses machine learning to select principal components as the set of weak classifiers and reinforces the adaptive boosting technology through an approach of central tendency. It can therefore avoid the over-fitting problem and solve the overlapping error trap of weak classifiers. The experiment exhibits that it can be effectively used in intelligent English essay assessing systems, with an overall 94% adjacent accuracy and all error rates less than 20% without singularity error in comparison to human scoring samples. The algorithm has a theoretical and practical value to intelligent scoring and machine learning.

Key words: intelligent essay evaluation, Adaboost/CT, indexes systems, hierarchical analysis