Visual Analysis Of Review Data Based On Topic Mining

With the growing popularity of the internet,there are an increasing number of user generated reviews on the Internet,which have become important information for consumers to make decisions and marketers to understand consumers' feedback.However,the amount of reviews is very large and the reading process is time consuming.Thus,how to dig useful information from reviews is a challenge.Although text miming is able to extract valuable information,such as topics and sentiments,from the text,it is not able to help users intuitively understand the topic distribution and relation of the text according to their personal preferences.Thus,in this paper,we propose two text mining based visualization systems to visually explore the topic geographical distribution of reviews.The topic geographical distribution system involves review text and spatial information to discover and explore topics from city level.First of all,the topic words which users may be interested in will be recommended.Secondly,the topic distribution is displayed by using the topic heat map and the topic time curve.Through the case study,the distribution patterns of different business in different cities are analyzed and the system can help users to make travel decision.In order to fully explore the topics of one business,we propose a visualization system based on text mining to present the differences of the topics between different businesses.Hierarchical clustering is applied to extract topics of reviews automatically.The hierarchy and similarity relationships between topics are displayed with hierarchical graph and bubble diagram and the word cloud is also incorporated to help users explore the topic context in depth.Finally,a case study is conducted to evaluate the effectiveness of the system.