| 基于BiLSTM-CRF的政府微博舆论观点抽取与焦点呈现 |
| |
| 引用本文： | 胡吉明,郑翔,程齐凯,张岩.基于BiLSTM-CRF的政府微博舆论观点抽取与焦点呈现[J].情报理论与实践,2021(1):174-179,137. |
| |
| 作者姓名： | 胡吉明 郑翔 程齐凯 张岩 |
| |
| 作者单位： | 武汉大学信息管理学院;武汉大学信息检索与知识挖掘研究所;武汉大学测绘遥感信息工程国家重点实验室 |
| |
| 基金项目： | 国家自然科学基金面上项目“基于深度学习的政务新媒体互动内容摘要自动生成与情感分析模型研究”的成果,项目编号:71874125。 |
| |
| 摘 要： | 目的/意义]准确把握公众微博评论中所反映的公众观点并总结舆论焦点,有助于及时获取和引导社会舆情态势,对政府公信力、快速响应能力及执行力提升具有支撑作用。方法/过程]文章针对当前政府微博评论社会功能发挥的现实要求和其文本特征挖掘的技术需求,从基于深度学习的文本智能语义理解和挖掘出发,提出了适用的细粒度四元组标注策略,构建了政府微博评论观点抽取与焦点呈现的深度学习模型POF-BiLSTM-CRF,即通过细粒度标注策略确定、Word2vec训练词向量、BiLSTM评论特征学习进行标签及其概率输出、CRF学习上下文实现微博评论标注优化,以及观点聚类和主题词提取后最终呈现舆论焦点。结果/结论]针对"中国警方在线"微博评论的实验表明,文章所提研究框架和模型能够有效进行舆论观点的智能化提取,为快速把握公众观点及为政府决策提供了参考。
|
| 关 键 词： | 政府微博评论 舆论观点抽取 深度学习模型 BiLSTM-CRF模型 POF-BiLSTM-CRF模型 |
Public Opinion Extraction and Focus Presentation in Government Microblog Based on BiLSTM-CRF |
| |
| Abstract: | Purpose/significance] Accurately understanding the public opinion reflected in comments of government microblog and summarizing the focus of public opinion will help to obtain and guide the situation of public opinion timely.It will also improve the government’s credibility,responsiveness and executive capability.Method/process] Based on the current requirements of social function exertion and text feature mining of government microblog comments,this paper proposes an applicable fine-grained quaternion annotation strategy and the POF-BiLSTM-CRF model to extract and present government microblog comments opinion,from the perspective of text intelligent semantic understanding and mining based on deep learning.In this paper,fine-grained annotation strategy is used to extract comment connotations;Word2 vec is used to train word vector;BiLSTM model is used to learn text features of Public microblog comments and output possible labels and probabilities;CRF model is used to optimize annotation by learn the context;clustering algorithm and topic words extraction algorithm are used to present the focus of public opinion.Result/conclusion] The experiment of comments of "China police online" shows that the research framework and model proposed in this paper can effectively solve the problem of public opinion extraction intelligently and rapidly,which provides a reference for the rapid grasp of public opinion and government decision-making. |
| |
| Keywords: | public microblog comments opinion extraction deep learning model BiLSTM-CRF model POF-BiLSTM-CRF model |
| 本文献已被 维普 等数据库收录！ |
|
