面向媒体信息传播的转载行为分析方法研究

随着互联网、大数据和新媒体的迅猛发展，如何利用先进的传播计算技术加强媒体传播能力建设至关重要。本课题面向媒体宣传领域的业务需求，以课题组构建的全球媒体传播分析系统为依托，借鉴深度神经网络在网络表示学习、文本表示学习等领域的研究成果，从媒体转载预测、新闻转载识别、转载群体分析三个角度展开对媒体转载行为的分析研究，解决媒体转载传播网络动态、稀疏、新闻转载模式复杂多样带来的信息传播分析困难的技术挑战，实现媒体信息的准确传播追踪。主要内容包括：基于注意力机制的媒体转载预测方法；基于多层次语义建模的新闻转载识别方法；基于BERT和变分图自编码模型的转载群体分析方法；在此基础上，研制一个媒体转载行为分析原型系统。本课题的研究对提升媒体传播能力有重要的理论意义和实际应用价值。

With the rapid development of the Internet, big data and new media, how to adopt advanced propagation computing technology to strengthen the propagation capacity becomes a very important and challenging research topic. In this project, based on the needs of media analysis,the constructed media propagation analysis system, and the latest techniques of network embedding learning and text representation learning,we will focus on analyzing the reprint patterns from three perspectives: media reprint prediction, news reprint identification, and reprint group analysis. The technical challenges caused by the dynamic and sparse reprint network and the complex reprint patterns will be solved, thus realizing accurate media information propagation. The primary contents of this project are as follows: media reprint prediction method based on attention mechanism, news reprint identification method based on multi-level semantic modeling, reprint group analysis method based on BERT and variational graph auto-encoder. Based on these new methods, we will design a novel prototype for media reprint patterns analysis. This research has important theoretical significance and practical application value for improving the media propagation ability.