社交媒体中情感传播关键问题研究

发布时间：2018-12-15 07:33

【摘要】：随着Web2.0的发展,用户通过社交媒体上传文字、图像、音频、视频等分享自己的状态或感兴趣的内容。用户的转发行为使得信息以指数级别的速度传播,这远远超过传统媒体的传播速度,因此社交媒体已渐渐取代传统的新闻媒体成为社会大众获取信息的重要渠道。社交媒体中最基本的是文字信息,其不仅包含实质性内容,还包含了用户对事物或事件的评价,评价表达了用户的情感。通过用户的转发行为,情感随着文字信息得到传播。根据心理学的情绪感染理论,情感极易在用户之间相互感染,引起社会大众的广泛关注和讨论。本文主要围绕社交媒体中的情感传播从四个方面进行深入的分析和研究。1.提出一种基于情感一致性的用户影响力排序算法用于寻找意见领袖。该算法主要研究在转发过程中情感与原情感保持一致的情况下产生的影响力。提出情感一致值的概念用于表示用户在转发过程中与原微博情感保持一致的程度,并提出情感一致权重的概念表示转发过程中两个用户之间情感保持一致的程度。根据用户的特点将用户分为三类:叶节点、出度为叶节点的用户和剩余节点。叶节点的影响力为0;出度为叶节点的用户的影响力主要来源于出度用户的情感一致值;剩余节点的影响力则来自于出度用户的影响力和出度用户的情感一致值。通过新浪微博数据集验证模型的有效性。2.提出一种基于情感变化的独立级联模型用于解决正向社会影响力最大化问题。研究社会影响力最大化的模型是建立在信息传播模型的实验成果基础之上,提出基于情感变化的信息传播模型解释影响力产生的过程。信息传播过程如下:在信息传播初始阶段,用户以一定概率从不知道信息到持有正向或负向情感。然后,随着网络中越来越多的用户参与到信息的转发中,用户之间相互影响,即用户会以一定的概率决定是否改变初始情感。当网络中不再有用户改变情感时计算正向影响力。信息传播模型通过真实的社交媒体数据验证模型的有效性。基于情感变化的独立级联模型应用到三个真实的网络中计算正向影响力。通过与现有算法进行对比,基于情感变化的独立级联模型能够得到最大的正向影响力。3.提出基于情感一致性的情感预测模型用于提高预测目标情感的准确度。该模型提出大众情感的概念用于表示评论中大众的情感;提出大众情感一致性的概念用于表示用户与大众情感一致的程度。根据基于大众情感一致性,将用户分为三类:独立用户、赞同用户和初始用户。独立用户表示与大众情感相关性小的用户,即大众情感一致性值较小的用户;赞同用户表示与大众情感一致的用户,即大众情感一致性值较大的用户;初始用户表示没有历史记录用于分析与大众情感相关程度的用户。基于情感一致性的情感预测算法结合了大众情感、个人情感、朋友情感和准朋友情感来预测目标情感。针对不同类型的用户,采用不同的情感组合进行情感预测。通过新浪微博数据集验证模型,实验结果表明该模型比现有模型性能更好。4.提出两个基于情感的信息传播模型用于研究不同情感在社交网络中的传播过程:基于情感的传染病模型和基于情感的独立级联模型。两个模型的区别在于情感传播过程中情感是否发生变化。基于情感的传染病模型中假设情感在传播过程中不发生变化,该模型提出基于情感的权重表示用户与用户之间对于某种情感的转发强度。基于情感的独立级联模型则假设情感在传播过程中发生变化。该模型将情感传播分为三个部分:第一,提出转发概率的概念研究包含情感的信息是否被转发;第二,使用机器学习算法研究情感被转发后是否会发生变化;第三,提出转化权重的概念研究情感如果在转发后发生变化,则转化后的情感。使用新浪微博数据集验证模型,实验结果表明基于情感的独立级联模型的性能更好。
[Abstract]:With the development of Web2.0, users can upload text, images, audio, video, and so on through social media to share their status or content of interest. The user's forwarding behavior makes the information spread at an index level, which is much higher than the propagation speed of the traditional media, so the social media has gradually replaced the traditional news media as an important channel for the public to obtain information. The most basic of the social media is the text message, which contains not only the substantive content, but also the user's evaluation of things or events, and the evaluation of the user's feelings. The emotion is spread with the text information through the user's forwarding behavior. According to the emotional infection theory of psychology, the emotion is very easy to be infected with each other among the users, which causes a wide range of attention and discussion among the public. This article mainly focuses on the deep analysis and study of the four aspects of the emotional communication in the social media. A user influence ranking algorithm based on emotional consistency is proposed to find the opinion leader. The algorithm mainly studies the influence of the emotion and the original emotion in the forwarding process. The concept of the emotion-consistent value is used to indicate the degree of the user's consistency with the original micro-emotion in the forwarding process, and the concept of the emotion-consistent weight is proposed to indicate the degree of the relationship between the two users in the forwarding process. according to the characteristics of the user, the user can be divided into three categories: leaf nodes, users of the leaf nodes and the remaining nodes. the influence of the leaf node is 0; the influence of the user with the degree of the leaf node is mainly from the emotion consistent value of the degree user; and the influence of the residual node is from the influence of the degree user and the emotion consistent value of the degree user. The validity of the model is verified through the sina microbo data set. An independent cascade model based on the change of emotion is proposed to solve the problem of maximizing the influence of the positive society. The model of maximizing the influence of social influence is based on the experimental results of the information transmission model, and the process of interpreting the influence based on the information transmission model based on the change of emotion is put forward. The information dissemination process is as follows: At the initial stage of the information dissemination, the user never knows the information to hold forward or negative feelings at a certain probability. Then, as more and more users in the network participate in the forwarding of information, the user interacts with each other, that is, the user decides whether to change the initial emotion with a certain probability. the positive influence is calculated when the user no longer has a user changing the emotion. The information dissemination model verifies the validity of the model through the real social media data. The independent cascade model based on the change of emotion is applied to three real networks to calculate the positive influence. By contrast with the existing algorithm, the independent cascade model based on the change of emotion can get the maximum positive influence. An emotion prediction model based on emotional consistency is proposed to improve the accuracy of the prediction target emotion. The model puts forward the concept of mass emotion to express the emotion of the general public in the comments, and puts forward the concept of the public sentiment consistency to show the degree of the user's agreement with the general public. The user can be divided into three categories based on the public emotional consistency: the independent user, the approval user and the initial user. The independent user represents a user with a small influence with the general public, that is, a user with a small general feeling consistency value, and a user who agrees with the general public feeling, that is, a user with a large emotional consistency value; the initial user indicates that there is no history of history to analyze the user associated with the general sentiment. The emotion prediction algorithm based on the emotion consistency combines the emotion of the public, the personal emotion, the friend emotion and the quasi-friend emotion to predict the target emotion. For different types of users, different emotion combinations are used for emotional prediction. The results show that the model is better than the existing model. Two emotion-based information transmission models are proposed to study the transmission process of different emotions in social networks: an emotional-based epidemic model and an emotion-based independent cascade model. The difference between the two models is whether the emotion changes in the process of emotional communication. In the model of the infectious disease based on emotion, it is assumed that the emotion does not change in the course of the transmission, and the model proposes that the weight based on the emotion indicates the forwarding strength between the user and the user for some kind of emotion. The emotion-based independent cascade model assumes that the emotion changes in the course of propagation. The model divides the emotion propagation into three parts: first, the concept of the forward probability is proposed to study whether the information containing the emotion is forwarded; secondly, whether the emotion is changed after the emotion is forwarded by using the machine learning algorithm; and thirdly, The concept of transformation weight is proposed to study the emotion if it changes after being forwarded. The results show that the performance of the independent cascade model based on emotion is better.
【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：G206;TP393.09

【相似文献】