基于用户评论的群体情绪识别与演化研究

发布时间：2018-12-14 08:31

【摘要】：随着电子商务的迅猛发展,网络购物已经成为当下中国最热的潮流。消费者在进行网络购物的同时,也会根据网络购物过程的体验和产品使用情况发表用户评论。每一条评论都是带有情感色彩的,是用户一种情绪的表达,或是喜欢、厌恶,或是对商家的意见和建议,准确及时地获知这些用户情绪对商家的经营决策至关重要。然而这些用户评论数量巨大且属非结构化信息,只关注每一条个体情绪是没有意义的,也是不可操作的。因此,需要将所有用户评论综合起来,研究用户群体情绪的变化,从而有效地支持商家的经营决策。本研究主要针对大型B2C电子商务网站上用户评论进行研究,使得这些评论信息可以有效指导商家的经营决策,具体研究内容如下:(1)文章首先对用户评论数据进行预处理,采用关联规则寻找频繁项集的方法实现了产品特征的抽取。根据已抽取到的产品特征,利用产品特征词和情感倾向词的共现关系实现了“产品特征—情感倾向”词对的抽取。然后,基于模糊数理论构建了情感倾向词模糊语料库,计算出了情感倾向词的模糊隶属度函数和情感极性值。并采用模糊认知图对“产品特征—情感倾向”词对及其情感极性值进行了知识表示,构建了“产品特征—情感倾向”综合分析模型;(2)在此基础上,利用已经抽取到的“产品特征—情感倾向”词对及其情感极性值构建了源案例库。基于证据理论对“产品特征—情感倾向”词对进行两两融合,实现了群体情绪的识别,并进行了群体情绪识别实验;(3)最后,文章基于时间序列实现了群体情绪的演化研究,追踪了群体情绪的变化过程和趋势,这是本文最大的创新点。并依据群体情绪演化结果,指导了商家的经营决策。本文把群体情绪的概念引入到用户评论挖掘中来,并基于时间序列对群体情绪的演化过程进行研究,在理论方面能够进一步丰富充实网络用户评论领域的研究;在实践方面,可以使商家准确及时地掌握用户的需求、喜好及变化趋势,更加高效的指导商家经营决策。
[Abstract]:With the rapid development of e-commerce, online shopping has become the hottest trend in China. While shopping online, consumers will also make comments according to the experience of online shopping process and the use of products. Each comment is emotional, it is the expression of a kind of user's emotion, or likes, dislikes, or to the merchant's opinion and suggestion, accurate and timely know these user's emotion is very important to the businessman's management decision. However, these users comment on a large number of unstructured information, focusing on each individual emotion is meaningless, and not operational. Therefore, it is necessary to synthesize all user reviews and study the changes of user group emotions, so as to effectively support business decisions. This research mainly focuses on the user comments on large B2C e-commerce websites, which can effectively guide the business decisions. The specific contents are as follows: (1) the paper preprocesses the user comment data. Product feature extraction is realized by using association rules to find frequent itemsets. According to the extracted product features, the cooccurrence relationship between product feature words and affective predisposition words is used to realize the extraction of "product-affective tendency" word pairs. Then, based on the fuzzy number theory, the fuzzy corpus of affective predisposition words is constructed, and the fuzzy membership function and affective polarity value of affective predisposition words are calculated. The fuzzy cognitive map is used to express the word pair of "product feature-affective propensity" and its affective polarity value, and a comprehensive analysis model of "product feature-affective propensity" is constructed. (2) on this basis, the source case database is constructed by using the extracted word pair of "product character-affective tendency" and its affective polarity value. Based on the evidence theory, this paper combines the word pairs of "product characteristics and affective tendency", realizes the recognition of group emotion, and carries on the experiment of group emotion recognition. (3) finally, based on the time series, this paper realizes the research of the evolution of group emotion, and tracks the changing process and trend of group emotion, which is the most innovative point of this paper. And according to the group emotion evolution result, has guided the merchant's management decision. In this paper, the concept of group emotion is introduced into user comment mining, and the evolution process of group emotion is studied based on time series, which can enrich the research of network user comment field in theory. In practice, the business can accurately and timely grasp the user's needs, preferences and trends, and guide business decisions more efficiently.
【学位授予单位】：东华大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F724.6;F713.55

【参考文献】