Research And Implementation Of Text Visualization Techniques For Social Network

With the development of Internet technology, social network is playing an increasingly important role in people's daily lives. Comparing to traditional blog and BBS, social network has faster spread speed, larger amount of users and more frequency interaction. Thus social network data is of great diversity and numerous, which makes it hard to obtain valuable information from it. Visualization is a technology that can display information vividly and intuitively. Besides, the interaction provided by it enables users to further explore the data after getting insight from the display. However, current visualization of social network mainly aims at emotional analysis, spam identification, communication analysis and text classification.Lacking of social network visualization clued by theme, users are hard to obtain interested information out of massive data. Therefore, theme-clued social network visualization is one of the most popular research fields.The method of social network visualization is studied in this thesis based on the above background, and the work is carried out in the following aspects:1. The technology of network data crawling is studied and the chosen website is analyzed.The network data crawling scheme is designed based on the problems encountered in data collecting, which combines Requests, Selenium-Driver and Python multithreading module threading, so as to accomplish efficient data crawling both of static webpages and dynamic webpages. The data crawled from network is stored in the MySQL table designed according to data characteristics.2. An interactive ThemeRiver is proposed to overcome the defect above. First, the rivers are sorted based on Coefficient of Variance. Second, the concept of ambiguity point is put forward,and the design and calculating method is elaborated. Third, an interactive method is provided to enable users to resort the river to eliminate the ambiguity. Experimental results validate the effectiveness of the proposed method in visualizing time-series text and eliminating the ambiguity of traditional ThemeRiver.3. A visualization method of data that is both grouped and ranked is proposed. The group information of each entity is visualized based on fan chart partition, and the rank of each entity within the group it belongs to is visualized by stratification layout. And the color, size and shape of each entity are designed. Experimental results show that the proposed method is able to make full use of space and visualize the grouped rank data effectively and intuitively.4. Data analysis and preprocessing is carried out based on the data crawled from social network, and a social network visualization tool is designed and implemented including six modules: topic structure module, interactive ThemeRiver module, word cloud module, users display module, questions display module and answers display module. Besides, interaction is designed between each module to achieve social network visualization clued by theme.