摘要
互联网短文本语言自由、灵活且缺乏规范性、要素错综复杂,使得传统的文本序列标注对情感要素抽取效果并不理想。针对此特点,文中提出一种基于长短时间记忆网络模型的互联网短文本情感要素抽取方法。主要利用长短时间记忆网络模型构建面向互联网短文本情感要素抽取任务的encode-decoder序列标注框架模型,并以此为基础融入3元窗口情感特征选择,在COAE2014测评数据集上实验。实验结果表明,该模型通过情感特征注入情感要素抽取准确率达70. 7%,利用浅层机器学习模型分析情感倾向性也取得了较好的效果。
Nowadays, the lnternet short texts are flexible, lack of normative characteristics and intricate elements. Therefore, the traditional text sequence is not ideal for the extraction of emotional elements. At this point, this article proposed a method based on the cyclic neural network model of Long Short Term Memory to extract the emotional elements of the Internet short text. The coding model of the encode - decoder sequence of the short elements of the Interact short essay was constructed by Long Short Term Memol7. The sentimental teature selection of 3 - element window was integrated and tested on the COAE2014 test dataset. The experimental results showed that the accuracy of the model extracted by emotional teatures was 70.7% , and the use of shallow machine learning model analysis for emotional propensity also achieved good results.
作者
尹光花
刘小明
张露
杨俊峰
YIN Guanghua;LIU Xiaoming;ZHANG Lu;YANG Junfeng(Department of Computer Science,Zhongyuan University of Technology,Zhengzhou 450007,China;Engineering Laboratory of Computer Information System Security Assessment,Zhengzhou 450007,China)
出处
《电子科技》
2018年第11期38-41,46,共5页
Electronic Science and Technology
基金
国家自然科学基金(61672361
U1404606)
河南省教育厅科学技术研究重点项目(14A520015)
关键词
互联网短文苓
文苓序列标注
长短时记忆网络模型
特征选择
要素抽取
机器学习模型
internet short text
text sequence annotation
long short term memory model
ieature selection
tea- ture extraction
machine learning model
