基于Harmonic函数的自动试题标注模型  

The Auto Test Question Annotation Model Based on Harmonic Function

作　　者：谢莹[1] 许荣斌[1] XIE Ying;XU Rong-bin(School of Computer Science and Technology,Anhui University,Hefei 230601,Anhui,China)

机构地区：[1]安徽大学计算机科学与技术学院,合肥安徽230601

出　　处：《韶关学院学报》2018年第6期1-6,共6页Journal of Shaoguan University

基　　金：国家自然科学基金(61602005);安徽省自然科学面上基金(1608085MF130)

摘　　要：现有的文本标注工作较少考虑关键字和类区域之间的相应关系,为减少在组卷过程中大量试题标注的人力成本,在充分研究试题数据集特征的基础上,将拉普拉斯映射稳定性与Harmonic函数属性相结合,构建试题数据相似度关联,正则化类群.对少量已加标注的试题样本进行训练学习,得到训练半监督分类模型,并用于大量未标注样本的自动标注工作.在2种数据集上进行3种方法的实验工作,实验结果较为显著.The relation between keywords and class area is ill considered by most extant methods of textual annotation. To reduce the labor costs of test question annotation when constructing test papers, which depend upon lots of studies on the features of test question data sets, we combine Laplacian embedding with Harmonic function＇s character to construct data similarity correlation for normalized class clusters. We have conducted training work on a few labeled test question samples and have produced a semi-supervised classification model. We use three methods in two datasets. Numerous experiments evidence that the annotation result is effective when this model is applied to a large number of unlabeled datasets.

关 键 词：试题标注 拉普拉斯映射 Harmonic函数 半监督分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]