基于内容挖掘的在线用户评论时间特征及其影响研究

发布时间：2017-12-31 15:05

  本文关键词：基于内容挖掘的在线用户评论时间特征及其影响研究　出处：《电子科技大学》2016年博士论文　论文类型：学位论文

  更多相关文章： 在线用户评论 时间特征 词向量表示 内容挖掘 动机

【摘要】：随着Web2.0的兴起,用户参与的观念日益深入人心。用户生成内容逐渐主导了网站内容的生成方式。用户生成内容中的在线用户评论是一种消费者自愿发布在公司或者第三方平台上的对产品或者服务的评估。在线用户评论对于消费者本身、潜在消费者、电子商务平台和生产厂商都起到了重要的作用。本文针对在线用户评论做了三方面的工作:首先,探索消费者在网上购物之后会在什么时候去网上发布他们的评论。其次,识别出消费者在这些评论中都发布了哪些方面的内容。最后,探索在线用户评论内容中隐含的动机和原因对消费者以这样的时间特征规律来发布在线评论的影响。针对消费者在购买以后什么时候去网上发布他们的评论这个问题,本文采用人类行为动力学的理论和方法来刻画了消费者在线购买和评论行为的时间特征规律。利用真实数据发现消费者“购买-评论”行为的时间间隔服从幂律分布。通过对消费者“购买-评价”时间间隔序列进行分段,可以发现在划分的不同阶段内评论者表现出了不同行为模式。针对消费者在评论中发布内容的问题。本文针对在线评论海量、非结构化、特征稀疏、以及通过传统挖掘方法获得结果语义理解差的问题,提出了基于词向量表示的在线用户评论内容挖掘方法。为了验证该方法的有效性,本文利用该方法在真实的在线用户评论数据集进行属性抽取和聚类;同时为了验证该方法的通用性,本文利用该方法对社交媒体中企业事件的识别。探索在线用户评论内容中隐含的动机和原因对消费者以这样的时间特征规律来发布在线评论的影响。首先是探索消费者以这样的时间特征规律发布在线评论的动机和原因,其实就是要找出哪些因素影响了消费者购买后发布评论的时机。本文首先结合消费者动机理论和社会交换理论,提出了一个从消费者自身动机、电商平台、产品体验和消费者社会关系四个方面来影响消费者评论及时性的理论框架模型。其次使用“评论-特征”挖掘和映射方法,从海量在线用户评论中抽取出理论模型中的变量。最后利用文本中抽取的特征和回归建模方法来验证提出假设。研究结果表明:会员等级、操作体验对评论及时性产生正向影响;支付价格、情感极性倾向、电商平台服务、物流服务、外观、操作系统和性价比对在线用户评论及时性产生负向影响。本文的主要贡献体现在以下三个方面:(1)系统刻画了电子商务环境下消费者“购买-评价”两种行为的活动规律。本文采用在线人类行为动力学中对行为规律的刻画方法,对电子商务环境下消费者“购买-评价”这两种典型行为的活动规律进行了系统刻画。验证了电子商务环境下消费者“购买-评价”行为之间的时间特征规律也服从幂律分布,从而为在线人类行为动力学的研究提供了新的实证结果。(2)提出了一种新的在线用户评论内容挖掘方法。针对海量在线用户评论中的特征稀疏、通过传统挖掘可理解性差的特点,本文提出了一种基于词向量表示的在线用户评论内容挖掘方法和一个半监督的“评论-特征”映射方法,解决了从海量稀疏短文本挖掘中的精度差和语义可理解性差的问题。通过利用该方法对在线用户评论的特征抽取和聚类,证实了方法的有效性;同时通过利用方法来对社会媒体中企业事件的识别,证明了该方法的通用性。(3)构建出在线用户评论及时性影响因素的理论模型本文采用消费者行为动机理论和社会交换理论,从消费者动机、产品、电子商务平台和消费者的社会关系这四个方面出发,构建出对在线用户评论及时性影响的理论模型。通过根据评论情感、会员等级以及通过文本挖掘的方法识别出在线评论内容中影响消费者评论发布时间的潜在因素来建立计量经济建模的方法,验证了各因素对在线用户评论及时性的影响关系。通过识别出的影响消费者评论行为的因素,电子商务企业对消费者进行响应更有针对性,从而获得有价值的在线用户评论和产生口碑。
[Abstract]:With the rise of Web2.0, user participation concept increasingly popular. Generation of user generated content gradually came to dominate the web content. The online user reviews of user generated content in a consumer is released voluntarily in the company or the third party platform of products or services. The online user reviews for evaluation of consumers, potential consumers, electronic business platform and manufacturers have played an important role. According to the online user reviews for three aspects: first, consumers will explore what time go online to post their comments after shopping online. Secondly, identify the consumers are released which contents in these comments. Finally, explore effect of implicit motives and reasons of online user reviews in the content of consumer characteristics such as time to publish online reviews for the consumer. For online publishing their comments on the issue after the purchase to what time, time characteristics this paper uses the theory and method of human behavior dynamics to describe and comment on the consumer online purchase behavior. Using real data found that consumers "buy - Comment" for the time interval obeys power-law distribution. Based on the consumer segment "buy - Evaluation" time interval sequence, can be found in the different stages of the reviewers showed different patterns of behavior. For the consumer publication in a review of the problem. According to the online reviews massive, unstructured, sparse features, and the semantic understanding of the problem of poor through traditional mining methods, and puts forward the method of mining online user reviews content words based on vector representation. In order to verify the validity of the method, this paper use the method in the real in Line user reviews data sets of attribute extraction and clustering; general at the same time in order to verify the method, the identification of enterprise social media events by using this method. To explore the impact of implicit motives and reasons of online user reviews in the content of consumers with time characteristics of law such to publish online reviews. The first is to explore the motivation and the reason consumers published by the characteristics of the time such online comments, in fact is to find out the factors which influence the consumer purchase after the release of time. This paper reviews the consumer motivation theory and social exchange theory, put forward a from consumers' motivation, business platform, the four aspects of social relations and consumer product experience effect of the theoretical model. Secondly, consumer reviews "comments - feature mining and mapping method, from massive online users Comment on the extracted theory of the variables in the model. Finally the characteristics and regression method of text extraction to verify the proposed hypothesis. The results show that: the grade of membership, operating experience has a positive impact on the timeliness of the price paid, comment; sentiment tendency, electronic business platform services, logistics services, appearance, price and operating system have a negative impact on the online user reviews. The main contributions of this paper are embodied in the following three aspects: (1) the system characterizes the consumers "under the environment of e-commerce purchase evaluation" two act law. This paper uses the method to characterize the behavior of online human behavior dynamics, activity patterns of consumers under the environment of e-commerce purchase evaluation "of the two typical behavior systematically characterized. Verify between consumers" under the environment of e-commerce purchase behavior evaluation " Time distribution also obeys the power-law distribution, which provides new empirical results for the study of human dynamics online. (2) proposed a method of online user reviews the contents of the new mining. According to the characteristics of sparse massive online user reviews, the traditional mining can understand characteristics of difference, is proposed in this paper. "Comments feature of an online user comments based on word vector mining method and 1.5 supervised mapping method, solves the sparse from the massive short text mining in semantic precision and poor intelligibility problems. The feature extraction and clustering using the method of online user reviews, proved to be effective methods; at the same time by using the method of corporate social media event recognition, demonstrate the versatility of the method. (3) to construct a theoretical model of factors influencing timeliness of online user reviews This paper uses consumer behavior motivation theory and social exchange theory, products from consumers, motivation, the four aspects of e-commerce platform and consumer society relationship, constructing a theoretical model of online user reviews. According to the review by emotion, method to establish econometric modeling potential factors influencing consumer review time release member grade and recognition through text mining method of online review content, verify the relationship between various factors of online user reviews. The review timeliness of consumer behavior factors identified, in response to more targeted e-commerce enterprises to consumers, so as to obtain the online user reviews of value and reputation.

【学位授予单位】：电子科技大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：F724.6;F274
，

本文编号：1360264