中文微博用户性别分类方法研究

该文旨在研究中文微博用户的性别分类问题,即根据微博提供的中文文本信息对注册用户的性别进行识别。虽然基于微博的性别分类已经有一定研究,但是针对中文的性别分类工作还很缺乏。该文首先提出分别利用用户名和微博文本构建两个分类器对用户的性别类型进行判别,并对不同的特征(例如,字特征、词特征等)进行了研究分析;其次,在针对用户名和微博文本的两个分类器的基础上,使用贝叶斯融合方法进行分类器融合,从而达到采用这两种文本分类信息同时对用户性别进行性别判断。实验结果表明该文的方法可以达到较高的识别准确率,并且分类器融合的方法明显优于仅利用用户名或者微博文本的分类方法。     

基于社交圈层和注意力机制的信息热度预测

社交网络现已成为现实世界中信息传播与扩散的主要媒介,对其中的热点信息进行建模和预测有着广泛的应用场景和商业价值,比如进行信息传播挖掘、广告推荐和用户行为分析等.目前的相关研究主要利用特征和时间序列进行建模,但是并没有考虑到社交网络中用户的社交圈层对于信息传播的作用.本文提出了一种基于社交圈层和注意力机制的热度预测模型SCAP(Social Circle and Attention based Popularity Prediction),首先对社交圈层进行定义,通过自动编码器提取用户历史文本序列的特征,对不同用户的社交圈层进行聚类划分,得到社交圈层特征.进而对于一条新发布的文本信息,通过长短期记忆网络与嵌入层提取其文本特征、用户特征和时序特征,并基于注意力机制,捕获到不同社交圈层对于该文本信息的影响程度,得到社交圈层注意力特征.最后将文本特征、用户特征、时序特征和社交圈层注意力特征进行特征融合,并通过两个全连接层进行建模学习,对社交信息的热度进行预测.在推特、微博和豆瓣等四个数据集上的实验结果表明,SCAP模型的预测表现相比于多个对比模型总体呈优,在不同数据集上均方误差(MSE)分别降低了0.017,0.022,0.021和0.031,F1分数分别提升0.034,0.021,0.034和0.025,能够较为准确地预测社交信息的热度.本文同时探究了不同实验参数对于模型的影响效果,如用户历史文本序列的数量、社交圈层的数量和时间序列的长度,最后验证了模型输入的各个特征和注意力机制的引入对于模型预测性能提升的有效性,在推特数据集中,引入社交圈层和注意力机制,模型的MSE指标分别降低了0.065和0.019.     

基于卷积神经网络的图文融合媒体情感预测

近年来,用户在社交媒体上越来越多地使用多媒体内容来分享经历和表达情绪。相比单独的文本和图像,融合文本和图像的多媒体内容能够更为充分地揭示用户的真实情感。针对单一文本或图像的情感不明显问题,提出了一种基于卷积神经网络(CNN)的图文融合媒体的情感分析方法。该方法融合图像特征与三个不同级别(词语级、短语级和句子级)的文本特征构建CNN模型,以分析比较不同层次的语义特征对情感预测的影响。在真实数据集上的实验结果表明,通过捕捉文本情感特征和图像情感特征之间的内部联系,可以更准确地实现对图文融合媒体情感的预测。     

不可见字符的文本安全隐藏算法研究

随着互联网以及移动通讯技术的快速发展,文本隐藏技术具有重要的研究意义以及应用价值。论文分析了现有文献对文本隐藏的相关研究,并以此为基础,结合哥德尔数学编码以及对称散列加密技术,提出了一种基于Unicode零宽度字符的文本安全隐藏算法以及相应的提取算法。该编码系统可以支持126个ASCII字符的隐藏,包括所有的英文字母和一些基本的标点符号。根据不可见性、嵌入容量、鲁棒性以及安全性验证算法的有效性,将实验结果与现有同类技术进行比较。实验表明,该算法能够在增大嵌入容量的同时,为用户之间传输保密文本消息时提供端到端的安全性。     

基于双通道LSTM模型的用户性别分类方法研究

微博用户性别分类旨在根据用户信息进行用户性别的识别。目前性别分类的相关研究主要针对单一类型的特征(文本特征或者社交特征)进行性别分类。与以往研究不同,文中提出了一种双通道LSTM(Long-Short Term Memory)模型,以充分结合文本特征(用户发表的微博文本)和社交特征(用户关注者的信息)进行用户性别分类方法的研究。首先,利用单通道LSTM模型分别学习两组文本特征,得到两种特征表示;然后,在神经网络中加入Merge层, 结合两种特征表示进行集成学习,以充分学习文本特征和社交特征之间的联系。实验结果表明,相对于传统的分类算法,双通道LSTM模型分类算法能够获得更好的用户性别分类效果。     

一种基于节点密度分割和标签传播的Web页面挖掘方法

获取Web页面中的重要内容如文本和链接,在许多Web挖掘研究领域有着重要的应用价值.目前针对该问题主要采用Web页面分割和区块识别的方法.但现有的方法将Web页面中重要文本和链接的识别视为两个相互独立的问题,这种做法忽略了Web页面中文本和链接的内在语义关系,同时降低了页面处理的效率.文中提出了一种Web页面重要内容挖掘的统一框架,该框架主要由3个部分组成:第一,先将Web页面转换为DOM树表示,然后采用节点密度熵为度量将DOM树分割为不同的页面块;第二,采用基于K最近邻标签传播的半监督方法自动扩展页面块训练集;第三,在扩展的页面块训练集上对SVM分类器进行训练,并用来对页面块进行分类.采用该框架可以将Web页面块区分为多种类型,并且该框架独立于Web页面的类型和布局.我们在真实的Web环境下进行了广泛的实验,实验结果表明了该方法的有效性.     

基于社交网络分析和XGBoost算法的互联网客户流失预测研究

针对互联网行业的客户流失预测问题,提出了一种社交网络分析和机器学习相结合的客户流失预测方法。考虑到社交活动对用户流失的影响,首先采用社交网络分析方法从用户社交数据中提取特征,然后使用XGBoost(Extreme Gradient Boosting)算法来进行客户流失预测,最后将该方法与其他机器学习算法(Logistic回归、支持向量机和随机森林)进行比较。实验结果表明,所提出的社交网络分析和XGBoost相结合的客户流失预测方法优于传统方法。     

数字博物馆中的图像盲水印技术*

针对数字博物馆中图像数量大、类型多的特点,提出一种适用于多种类型图像的盲水印算法。该算法将用户ID作为水印信息,从而实现对图像的操作跟踪。水印嵌入时,首先将水印信息转换成ASCII码形式,并对其进行交织和BCH编码以提高抗突发和随机错误的能力,然后根据图像类型取得嵌入水印的载体,再对该载体语义丰富的位置进行分块DCT变换,最后通过修改子块DCT系数间关系嵌入水印。水印检测时,以嵌入过程的逆过程抽取水印信息,并计算其与所有用户ID的相似度来获得初始嵌入的ID。实验表明,该算法能抵抗一般的信号处理操作和特定条件下的几何变换。     

融合短文本层级注意力和时间信息的推荐方法

信息过载造成的数据稀疏性问题制约着基于评分数据的矩阵分解模型的推荐性能,融合评论文本的推荐模型能够有效缓解评分数据稀疏性.当前的推荐系统利用评论文本为用户和项目建模时,大多仅将用户对项目的评论作为数据来源,而忽视了时间信息对用户和项目属性的影响.针对此问题,提出了一种融合短文本层级注意力和时间信息的推荐方法(RHATR),该方法能够充分地挖掘评论文本潜在的语义信息,并为用户偏好和项目特征的动态变化进行建模.通过对单条评论文本应用单词级注意力,挖掘单条评论文本中情感词和关键词等有效信息,学习用户和项目表示;对含有时间因素的用户评论集和项目评论集分别应用评论级注意力,提取有效的评论,进一步学习用户偏好和项目特征动态表示.将从评论文本中学到的用户和项目表示以及基于ID的项目和用户嵌入作为最终特征,来捕获各用户和项目的潜在因素.实验结果表明,提出的方法相对于当前基线方法在Amazon和Yelp数据集上的均方根误差(RMSE)取得了较好的效果.     

基于层次型文本交互的表示学习研究

传统的推荐系统面临着诸如数据稀疏性、无法解释的推荐等几个挑战。为了解决这些问题,许多研究通过挖掘评论文本语义信息来提高推荐性能。然而,这些方法在文本特征建模和文本交互方面存在问题。在文本建模方面,它们简单地将用户/物品的所有评论拼接成一个单一的评论。然而,单词/短语级别的语义信息可能与评论文本的整体语义信息相悖。在文本交互方面,它们将交互推迟到预测层,无法捕捉用户和物品之间复杂的相关性。为了解决这些问题,我们提出了一种新颖的基于层次型文本交互的表示学习方法。在该方法中,我们以层级方式对低级单词语义和高级评论文本进行建模,以便在不同粒度上挖掘文本信息。为了进一步捕捉复杂的用户-物品的交互关系,我们提出在不同层次上挖掘用户-物品之间的语义关联。在单词级别上,我们提出了一种针对每对用户-物品个性化的注意力机制,来捕捉表示每个评论的重要单词。在文本级别上,我们在用户和物品之间相互传播文本语义信息,并捕捉针对目标任务有用的评论文本。最后,我们通过协同过滤框架,将该方法应用于评分预测应用场景,并通过在公开数据集上的对比实验,证明该方法在评分预测方面的性能优于现有方法。     

The reciprocal effects of social network site use and the disposition for self-disclosure: A longitudinal study

Since the advent of social network sites (SNSs), scholars have critically discussed the psychological and societal implication of online self-disclosure. Does Facebook change our willingness to disclose personal information? The present study proposes that the use of SNSs and the psychological disposition for self-disclosure interact reciprocally: Individuals with a stronger disposition show a higher tendency to use SNSs (selection effect). At the same time, frequent SNS use increases the wish to self-disclose online, because self-disclosing behaviors are reinforced through social capital within the SNS environment (socialization effect). In a longitudinal panel study, 488 users of SNSs were surveyed twice in a 6 months interval. Data were analyzed using structure equation modeling. The proposed reciprocal effects of SNS activities and self-disclosure were supported by the data: The disposition for online self-disclosure had a positive longitudinal effect on SNS use which in turn positively influenced the disposition for online self-disclosure. Both effects were moderated by the amount of social capital users received as a consequence of their SNS use.     

Social facilitation and human–computer interaction

The Computers As Social Actors (CASA) research paradigm has examined how individuals respond to computers programmed to interact in various ways. In the current research, we extend the principles of CASA to determine whether computer icons can be used to produce social facilitation effects. Varying task difficulty and the presence or absence of a computer icon (i.e., Microsoft word’s Clip), performance on a typing task is considered. Overall, results provide some support for the contention that the mere presence of a computer icon may influence task performance.     

Supporting controversial CSCL discussions with augmented group awareness tools

This paper describes the development of augmented group awareness tools that take mutual user ratings of their online discussion contributions as input, aggregate these data, and visually feed these data back to the members in real time, thereby informing participants about how the group as a whole perceives their contributions. A specific group awareness tool was experimentally tested in a CSCL scenario using online controversies about a physics domain. The learning material was distributed across group members to create a situation where an individual minority member with a scientifically correct viewpoint faces a majority favoring a plausible, but incorrect viewpoint. It was hypothesized that in unsupported CSCL groups an incorrect majority would dominate a correct minority, whereas in groups that were supported by an augmented group awareness tool minority influence could be strengthened by making minority contributions salient. The paper reports results in support of this hypothesis, and discusses the mechanisms leading to the benefits of group awareness tools for collaborative learning.     

基于Android平台的大学校园社交应用的设计与实现

目前的社交应用大部分都是面向所有用户进行设计和实现的．但这种方式并不能满足大学生这样的特殊用户群体的需求,为了方便大学生根据自己的需求在校同内进行社交活动,基于Android设计和实现一个便捷的社交应用平台,该平台能够方便大学生随时随地搜索和发起任何自己感兴趣的校园活动并参与其中．通过感兴趣的校园活动积累经验,提升自我价值。     

Social responses to Facebook status updates: The role of extraversion and social anxiety

Posting, reading, and responding to status updates is an integral part of many peoples' daily lives. However, the role of personality in predicting social responses to status updates remains largely unexplored. Based on the social enhancement and the social compensation hypothesis, we assessed the role of extraversion and social anxiety in predicting social responses to status updates. Moreover, we explored the influence of valence in this context. Capitalizing on the assets of a multimethod approach, personality was assessed with self-reports, and valence was evaluated by independent raters. Social responses to status updates were captured (1) by observing direct social feedback (i.e. likes and commenters) and (2) by informant reports on the interpersonal appraisal of participants’ status updates by their friends. In a German and a US sample, for direct social feedback neither extraversion nor social anxiety emerged as significant predictors. However, analyses of the informant reports showed that status updates of more socially anxious individuals were appreciated more by their friends. Furthermore, results pointed to the importance of valence in this context; revealing associations between valence and direct social feedback, valence and extraversion, and a moderation effect of personality on the association between valence and likes in the US sample.     

Types of humor that robots can play

Although humor is a well-known social lubricant defusing a complicated conflict between two parties, the efficacy of humor in human–robot interaction has barely been tested yet. This study compared the characteristics of humor performed by a robot and human to identify the possible type of jokes that a robot may play.In the experiment, a human actor performed disparaging – racist and sexist jokes, and non-disparaging (human condition and sexual) jokes, and a robot counterpart mimicked the same performance. Fifty-eight university students, 30 male and 28 female with mean age 23.10 (SD = 2.00), watched the randomly assigned jokes performed either by the robot or the human actor. The participants rated perceived humorousness, offensiveness, and willingness to share the joke with others, the perceived social presence and social attractions of the actor. The result showed that participants perceived non-disparaging jokes to be more humorous when performed by the human actor. On the other hand, the participants exhibited less disgust toward disparaging jokes when they were performed by the robot actor. This shows that humor can be used as an effective way to enrich the interaction between human and robot; but the acceptable types of humor should be carefully selected.     

Social Media and Trust during the Gezi Protests in Turkey

The Gezi Protests, an environmental sit‐in that turned into a social movement in Turkey, is often compared to the Arab Spring and the Occupy movement with regard to the importance attributed to social media. This paper examines the role that social media played during the protests, with an emphasis on how trust was built and maintained among the protestors. In‐depth interviews with 21 active Gezi protestors revealed that social trust and system trust were intertwined in actual practices. On one side, technological affordances worked as an interface that facilitated social identification, which helped in trusting the person behind the information. On the other side, technological affordances themselves invited different levels of trust, subject to both physical constraints and technological barriers.     

Folksonomy-based personalized search and ranking in social media services

In recent years, social Web users have been overwhelmed by the huge numbers of social media available. Consequentially, users have trouble finding social media suited to their needs. To help such users retrieve useful social media content, we propose a new model of tag-based personalized searches to enhance not only retrieval accuracy but also retrieval coverage. By leveraging social tagging as a preference indicator, we build two models: (i) a latent tag preference model that reflects how a certain user has assigned tags similar to a given tag and (ii) a latent tag annotation model that captures how users have tagged a certain tag to resources similar to a given resource. We then seamlessly map the tags onto items, depending on an individual user's query, to find the most desirable content relevant to the user's needs. Experimental results demonstrate that the proposed method significantly outperforms the state-of-the art algorithms and show our method's feasibility for personalized searches in social media services.     

Facebook: Social uses and anxiety

Research has shown inconsistent relationships between social anxiety and time spent on Facebook, possibly because Facebook’s many activities vary in degree of social interactivity. We examined the relationships between social anxiety, anxiety on Facebook, and social Facebook use. A multiple regression predicting social Facebook use revealed an interaction. Participants with high anxiety on Facebook and high social anxiety reported more frequent social Facebook use than those with high anxiety on Facebook and low social anxiety. A second multiple regression predicting social anxiety showed a suppression effect, indicating that social Facebook use predicts social anxiety only once anxiety on Facebook has been accounted for. These findings suggest that anxiety on Facebook clarifies the relationship between social anxiety and social Facebook use.     

How do people compare themselves with others on social network sites?: The case of Facebook

The social comparison theory and its subsequent studies say that comparing with others can influence an individual in several ways (e.g., evaluation of oneself, influence on self-esteem/self-confidence, and efficient decision making) and people compare with others when they are confronted with information of others. With the popularity of social network sites, many people acquire or are exposed to information of others on social network sites, which implies that people are likely to frequently engage in social comparison behavior on social network sites. The present paper examines social comparison behavior on social network sites (especially on Facebook) using a college students sample. We find that an individual’s personality characteristics (i.e., social comparison orientation, self-esteem, self-uncertainty, and self-consciousness) influence the person’s social comparison frequency on Facebook. A positive relationship between Facebook use intensity and social comparison frequency on Facebook is found. In addition, we find a positive association between social comparison frequency on Facebook and the frequency of having a negative feeling from comparison. Other findings are also reported in the paper.