Interactive Exploration Of Multi-faceted Relationships In Large Document Collections

With the rapid development of science and information technology,human society produces increasingly sharp text data in daily life.The corpus of a large amount of text information in any particular field often exceeds the scope that one can easily observe and analyze.Document collections not only contain rich semantic content,but also various text relationships,so the analysis of document collections is particularly important.Some existing methods of text analysis mainly focus on semantic content or a single kind of text relationship,and cannot treat the document collection as a whole to systematically and comprehensively explore from content to multiple text relationships.This paper defines a complete set of document relationships and implements modelling construction,at the same time,it proposes a visualization method to interactively explore the text content of the document collection and the comprehensive document relationships.The method mainly uses an improved multi-dimensional scaling technique to encode document relationships by projecting multiple document nodes.Each node uses word cloud to encode high-frequency vocabulary in the document,and processes the color of the words in the word cloud so that the colors have corresponding meanings.The tool integrates an overview view of content and document relationships and three auxiliary component views to support the viewing and exploration of interactive document word cloud nodes,the adjustment of document relationships,and the observation of actual relationships such as references.This paper demonstrates the effectiveness and usability of the method through a case study of a real paper data collection and two experimental evaluations.