基于主题挖掘的评论数据可视分析

发布时间：2018-10-10 18:40

【摘要】：随着互联网的普及,越来越多的消费者倾向于在网络上对商家进行评论,这些评论数据给出消费者对商家产品、服务等方面的评价信息,是商家获取消费者反馈的重要来源,也能帮助消费者选择合适的商家。随着评论文本越来越多,如何从海量评论文本数据中挖掘有价值的信息成为一个大挑战。文本挖掘技术从文本中发现主题及其用户评价,但无法直观地理解和探索主题的空间分布、主题的相互关联关系等。针对这一问题,本文提出了两个基于主题挖掘的可视分析系统。为了探索城市主题的地理空间分布,本文提出了城市主题地图系统,将商家的文本内容结合地理空间信息,进行城市规模的主题发现和探索。用户首先选择推荐的相似主题词,自主确定感兴趣的主题,其次通过主题热力图和主题时间曲线等可视化,探索商家在该主题下的城市空间和时间分布,并通过具体案例分析了不同城市不同主题的时空分布规律,有助于用户的商家决策。为了全面分析某个商家的所有主题,本文提出了基于意见挖掘的评论主题可视分析系统,展示不同商家的主题差异性。商家评论文本采用层次聚类的方式进行自动主题挖掘,并使用层次树图和气泡图来表达主题之间层次和相似关系,并辅助词云等方式帮助用户深入挖掘主题内容,最后通过超市和餐厅等案例,分析商家的主题分布及用户评价,验证了系统的有效性。
[Abstract]:With the popularity of the Internet, more and more consumers tend to comment on the business on the network. These comments data give consumers' evaluation information on products and services, which is an important source of consumer feedback. It can also help consumers choose the right business. With the increasing number of comment texts, how to mine valuable information from large amount of comment text data becomes a great challenge. Text mining technology can not directly understand and explore the spatial distribution of themes and the relationship between themes and so on. To solve this problem, this paper proposes two visual analysis systems based on topic mining. In order to explore the geographical spatial distribution of urban theme, this paper proposes a city theme map system, which combines the text content of business with geospatial information to discover and explore the theme of urban scale. The user first selects the recommended similar theme words, independently determines the topic of interest, and then explores the city space and time distribution of the merchant under the theme by visualizing the thermal map of the theme and the thematic time curve, etc. The temporal and spatial distribution of different themes in different cities is analyzed through specific cases, which is helpful to the decision making of users. In order to analyze all the topics of a merchant, this paper proposes a visual analysis system based on opinion mining to show the differences of different business themes. Merchant comment text is automatically mined by hierarchical clustering, hierarchical tree map and bubble graph are used to express the hierarchical and similar relationship between topics, and auxiliary word clouds are used to help users dig into the topic content in depth. Finally, through the case of supermarket and restaurant, this paper analyzes the distribution of business themes and user evaluation, and verifies the effectiveness of the system.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1

【相似文献】