In-depth Interactive Visual Exploration for Bridging Unstructured and Structured Document Content

AXEL J. SOTO; RYAN KIROS; VLADO KESELJ; EVANGELOS MILIOS

Philadelphia

Workshop; SIAM DM 2014 Workshop on Exploratory Data Analysis; 2014

Semi-structured data refers to the combination of unstructured and structured data. Unstructured data is free text in natural language, while structured data is typically stored in tables and following a data schema. Recent statistics shows that 80% of the data generated in the last two years is unstructured. However, one interesting observation is that free text usually comes along with some structured data, or meta-data describing or adding more information about the text. In this paper we present ViTA-SSD, a Visual Text Analytics Tool for Semi-Structured Data. This tool aims at extracting interesting patterns in semi-structured data through the joint consideration of the free text and the meta-data. This represents a challenging task because an effective approach needs the combined effort of text mining algorithms and human experts who can drive the exploration process in a meaningful way. A related challenge is the appropriate visualization and understanding of the patterns found. In order to address these challenges, our visual analytics tool takes advantage of a novel dimensionality reduction and a fast user-supervised clustering method. We showcase our tool here as well as we reflect on some lessons learned from the development and evaluation of our tool.