Visualization of Large Document Collections.
详细信息
文摘
With recent developments in content creation and storage,it is possible to generate,retrieve,and save information at a much faster rate. This information comes in a variety of forms and structures. Text is one important class of information. News articles,blog posts,emails,cell phone messages,and posts from social networking communities such as Facebook and Twitter are common. Unfortunately,reading text documents is still the predominant method used to understand them. In many situations it would be advantageous to have a more efficient and effective approach to assist readers in obtaining insights about large collections of text information. This dissertation describes a method to organize,visualize,and interact with large document collections. The goal is to provide visualization and interaction techniques that support efficient exploration and discovery within a set of documents. Rather than trying to abstract and present the text of the documents alone,we use a perception-based multidimensional visualization to show multiple document properties like document size,date of publication,grade reading level,and emotional affect. Documents with similar topics are spatially clustered into document stacks. Short text summaries are generated to describe the contents of the documents in a stack. A user interface is provided to allow viewers to interact with stacks and the documents they contain. Documents can be added,removed,or shifted between stacks to construct new arrangements based on a viewers domain expertise. We demonstrate our system using four collections of documents ranging in size and composition from a few dozen web pages to 10,000 newspaper articles.