摘要
新闻报道中观点能够影响读者的感受,针对目前新闻报道中观点提取缺失的现状。本文提出一种条件随机场(CRF)和深度学习相结合的模型,通过集成深度学习的BiLSTM方法和改进型CRF方法,实现对新闻文章的观点持有者、评价对象和观点极性3种实体信息的提取。试验表明:相较于CRF算法,准确率、召回率和F1值平均提高12.29%、10.00%和11.07%。
Opinions in news reports can affect readers’ opinions,aiming at the current situation of the lack of opinion extraction in news reports. This paper designs a model based on CRF and the combination of deep learning. By integrating the BiLSTM method of deep learning and the improved CRF method,three kinds of entity information:opinion holder,opinion target and opinion polarity,are realized for news articles. Compared with CRF algorithm,he precision,recall and F-measure are increased by12.29%,10.00% and 11.07% on average.
作者
周星瀚
刘宇
邱秀连
ZHOU Xing-han;LIU Yu;QIU Xiu-lian(Wuhan Research Institute of Posts and Telecommunications,Wuhan 430000,China;Nanjing Fiberhome World Communication Technology Co.Ltd.,Nanjing 210000,China)
出处
《电子设计工程》
2020年第3期18-22,共5页
Electronic Design Engineering
关键词
深度学习
双向长短时记忆网络
条件随机场
观点提取
新闻舆论
deep learning
Bidirectional Long Short-Term Memory(BiLSTM)network
Conditional Random Field(CRF)
opinion extraction
news media
