Text Visual Analytics Based On Topic Concurrence Relation

In the light of text mining, text semantics can not only be represented with a com-pounded semantic expression, but also be a combination of multiple topics. Current text visual analysis techniques represent text semantics based on either a compounded seman-tic expression, or relations between single topics and text semantics. Different from these two kinds of text semantic information, text visual analysis method in this thesis focuses on how users can effectively understand text semantics with the help of topic-topic concur-rence relations from the fact that text semantics can be described as topic distributions. Our research motivation is:Multiple topics to describe semantics can not only help users learn about semantics in detail, but also help users intuitively understand semantic differences among documents; Besides, topic concurrence relations can help users globally master text semantic structures. In order to verify the feasibility of text semantic understanding with the guide of topic concurrence relations, we take single long documents and huge text corpus for example, and propose appropriate visual design respectively to represent topic concur-rences to users. Experiments verify topic concurrence relations can satisfy our research motivation.