【作者】 阳曦； 柯登峰； 丁鹏； 徐波；

【Author】 YANG Xi1, KE Dengfeng1, DING Peng1, XU Bo1, 2 (1. Digital Content Technology Research Center, Institute of Automation, Chinese Academy of Sciences, Beijing 100080, China;2. National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sci-ences, Beijing 100080, China)

【机构】 中国科学院自动化研究所数字内容技术研究中心；

【摘要】 本文分别采用了多元线性回归算法和反向传播算法对二语学习中英语口语的人工评分和三种机器评分之间的关系进行学习,实现了上述三种机器评分的融合,并从语音库、相应的人工评分和机器评分三个角度对数据集的建立进行了详细的介绍。实验结果表明,由以上两种融合方法得到的机器总分与人工评分之间的相关度相比于融合前的最佳机器评分在句子层次上分别提高了1.4%和1.7%,在说话人层次上提高了0.6%,并显著降低了两者之间的误差平方和,证明了融合的可行性和有效性。更多还原

【Abstract】 The paper studied the relationship between the human scores and three machine scores of spoken English in second language learning by multiple linear regression and back propagation respectively and carried out the fusion of those three machine scores. It introduced the establishment of the database at three angles: speech corpus, the corresponding human scores and machine scores in detail. Results obtained from the experiments show that overall scores fused by two fusion methods mentioned above improve the correlation between machine scores and human scores by 1.4% and 1.7% in sentence level and 0.6% in speaker level compared with the best machine scores before fusion, and decrease the square sums of residual errors between them obviously. It proves the feasibility and validity of fusion.更多还原