最大似然估计真值推断及其在教学质量评价中的应用

MLE ground truth inference and its application in teaching quality evaluation

教学质量评估的一个常见手段是组织学生对于课程进行评分.由于学生普遍缺乏教育领域专业知识同时有具备非常强烈的主观偏好,因此传统的方法无法提供更加丰富的信息且估计的结果也不准确.论文提出一种新颖的教学质量评估框架.该框架首次将基于最大似然估计的Dawid&Skene算法应用到课程教学质量评估任务中,并对算法的实现过程进行改进和扩展,从而提高其性能并且支持从信息提取、教学质量评估以至低质量评估者过滤等一整套评估流程.实验结果显示论文提出的方法相对于传统方法能够提供更加丰富的信息并且其评估质量也更加可靠.

A common method used for evaluating teaching quality of a course is that students are arranged to provide scores to rank the quality of the course.Due to students ＇ lack of domain expertise and having strong individual preferences,traditional methods cannot provide enough information and the evaluation results are inaccurate.This paper put forward a novel teaching quality evaluation framework.The proposed framework first applied the maximum likelihood estimation （MLE） based on Dawid ＆ Skene algorithm to teaching quality evaluation,improved its implementation and extend it to the whole procedures of evaluation which included information extraction,teaching quality evaluation and low quality rater filtering.Experimental results showed that compared with traditional method the proposed framework provides more plentiful information with higher accuracy.