Visual Mining of Text Collections

Abstract

What happens if you have to examine and reach conclusions from a considerable number of textual documents? If you are faced with this task or with developing tools for completing this task, this tutorial is for you. Examining text is crucial for many different types of applications. Even applications that rely on additional types of data (such as images, signals, simulations) usually have complementary or alternative text based output. The challenge of interpreting content and extracting useful information from a document collection is the target of efforts in various areas of computer science. Fields such as Text Mining try to extract knowledge automatically or semi-automatically from collected textual information; however, due to the multi-dimensional characteristics of text it is paramount to couple these algorithms with meaningful visual representations in order to improve performance and allow the discovery of relevant information within a text data set. Since it is not feasible to go through the entire documents' content in detail due to data sizes and time constraints, Visual Text Mining (VTM) - the combination of Text Mining and Visualization - is focused on developing tools to help users extract meaning from text collections without extensive reading. In this tutorial we introduce the necessary background and the graphical techniques involved in Visual Text Mining of document collections.