查询结果:   刘杰,孙娜,袁克柔,余笑岩,骆力明.中文作文句间逻辑合理性智能判别方法研究[J].计算机应用与软件,2019,36(1):71 - 77.
中文标题
中文作文句间逻辑合理性智能判别方法研究
发表栏目
应用技术与研究
摘要点击数
992
英文标题
INTELLIGENT DISCRIMINANT METHOD OF LOGICAL RATIONALITY BETWEEN SENTENCES IN CHINESS COMPOSITION
作 者
刘杰 孙娜 袁克柔 余笑岩 骆力明 Liu Jie Sun Na Yuan Kerou Yu Xiaoyan Luo Liming
作者单位
首都师范大学信息工程学院 北京 100048     
英文单位
College of Information and Engineering, Capital Normal University, Beijing 100048, China     
关键词
作文评测 BiRNN 句间逻辑合理性 无监督学习 文本分类
Keywords
Composition evaluation BiRNN Logical rationality between sentences Unsupervised learning Text classification
基金项目
国家自然科学基金项目(61371194,61672361)
作者资料
刘杰,教授,主研领域:自然语言处理,语义网。孙娜,硕士生。袁克柔,硕士生。余笑岩,本科生。骆力明,教授。 。
文章摘要
在作文评测中,句间逻辑合理性是评价语言运用能力的一项重要指标。从文本分类的角度,对作文段落句间逻辑合理性进行定性分析。依据逻辑合理的段落其句子的位置是相对固定的,将现有的基于传统、基于深度学习的文本分类算法应用在中小学人物类作文段落句间逻辑合理性的判别上,实验结果表明使用分类模型对段落句间逻辑合理性判别是有效的。在此基础上,进一步提出代表句子特征的关键词筛选方法,实验结果表明使用句首关键词、句尾关键词或两者结合作为句子特征的方法,比句子全部关键词更能代表句子信息,明显提高判别的准确率。
Abstract
In composition evaluation, logical rationality between sentences is an important indicator to evaluate the ability of language application. From the perspective of text classification, this paper made a qualitative analysis of the logical rationality between sentences of paragraphs. According to the fact that the position of the sentence was relatively fixed in a logical and reasonable paragraph, we applied the existing text classification algorithms based on traditional and deep learning into the discrimination of logical rationality between sentences of paragraphs in character composition in primary and secondary schools. The experimental result shows that it is effective to use the classification model to distinguish logical rationality between sentences of paragraphs. On this basis, the keyword selection method representing the features of the sentence was further proposed. The experimental result shows that the method of using sentence heading keywords, sentence ending keywords or the combination of the two as the sentence features can represent the sentence information better than all the sentence keywords, which significantly improves the accuracy of discrimination.
下载PDF全文