MiTexCube: MicroTextCluster Cube for online analysis of text cells and its applications. (5th October 2012)

• Record Type: Journal Article
• Authors: Zhang, Duo; Zhai, ChengXiang; Han, Jiawei
• Is Part Of: Statistical analysis and data mining. Volume 6:Number 3(2013)
• Page Start: 243
• Page End: 259
• Publication Date: 2012-10-05
• Subjects: Data mining -- Statistical methods -- Periodicals; 006.312
• DOI: 10.1002/sam.11159
• Languages: English
• ISSNs: 1932-1864

Abstract

<abstract abstract-type="main" xml:lang="en"> <title>Abstract</title> <p>A fundamental problem of multidimensional text database analysis is efficient and effective support of various kinds of online applications, such as summarizing the content of a text cell or comparing the contents across multiple text cells. In this paper, we propose a new infrastructure called MicroTextCluster Cube (or <italic>MiTexCube</italic>) to support efficient online text analysis on multidimensional text databases by introducing <italic>micro‐clusters</italic> of text documents as a compact representation of text content. Experimental results on real multidimensional text databases show that (i) <italic>MiTexCube</italic> can be materialized efficiently with reasonable overhead in space, and (ii) applications based on the proposed materialized <italic>MiTexCube</italic> are more efficient than the baseline method of direct analysis based on document units in each cell, without sacrificing much quality of analysis, and <italic>MiTexCube</italic> naturally accommodates flexible trade‐off between efficiency and quality of analysis. © 2012 Wiley Periodicals, Inc. Statistical Analysis and Data Mining 6: 243–259, 2013</p> </abstract>