摘要
[目的/意义]挖掘网民舆情偏好,推荐网民感兴趣的舆情事件,解决“信息过载”背景下网民信息筛选难题、提高信息获取效率;提升平台用户体验,增加用户黏度。[方法/过程]基于NRL理论和思想构建EGE推荐模型,挖掘网民舆情偏好、推荐舆情事件。首先收集、预处理数据生成舆情事件共现网络;然后运用NRL相关算法得到舆情事件的低维向量表示,用高斯分布函数和已访问事件低维向量表示反映网民偏好,融入softmax与负采样以降低复杂度;最后对网民未关注的事件打分,运用KNN算法得到高评分事件集合TOP-M。加入当期其它类别的高关注度舆情事件形成最终的推荐列表。[结果/结论]基于网民历史访问记录运用舆情事件EGE推荐模型,能够有效地预测并推荐满足网民兴趣偏好的事件。
[Purpose/Significance]This research aims at exploring the lyric preferences of netizens,recommending lyric events that meet their preferences,solving the problem of screening information of netizens in the context of"information overload",improving the efficiency of information acquisition by netizens,and improving user's experience of the platform and increasing user retention.[Method/Process]Based on the theory and ideas of NRL,this paper constructs an EGE model to mine the preferences of netizens and recommend public opinion events.Firstly,the data is collected and preprocessed and the public opinion network is generated.Then the NRL correlation algorithm is used to obtain the public opinion events and the low-dimensional vector representation of the netizens.The Gaussian distribution function and the event low-dimensional vector representation are used to reflect the preferences of netizens,and they are integrated into softmax and Negative Sampling to reduce the complexity.Finally,the events that the netizens are not concerned about are scored,and the KNN algorithmn is used to get high score event list TOP-M.The final list of recommendations is formed with other categories of high-profile public opinion events in the current period being taken into consideration.[Result/Conclusion]Based on the historical access records of netizens,we can use the EGE public opinion recommendation model to predict events that satisfy the interest preferences of netizens.
作者
田世海
董月文
王健
Tian Shihai;Dong Yuewen;Wang Jian(School of Economics and Management,Harbin University of Science and Technology,Harbin 150040;School of Mechanical&Power Engineering,Harbin University of Science and Technology,Harbin 150040)
出处
《情报杂志》
CSSCI
北大核心
2020年第2期108-115,共8页
Journal of Intelligence
基金
黑龙江省自然科学基金项目“融媒体时代突发事件网络舆情引导机制研究”(编号:LH2019G017)
黑龙江省社会科学研究规划项目“黑龙江省大数据产业联盟云服务模式研究”(编号:16GLB01)研究成果之一
关键词
舆情偏好
舆情事件推荐
网络表示学习(NRL)
EGE推荐模型
lyric preference
public opinion event recommendation
Network Representation Learning(NRL)
EGE recommendation model
