TU Darmstadt / ULB / TUbiblio

Analytical Semantics Visualization for Discovering Latent Signals in Large Text Collections

Stab, Christian and Breyer, Matthias and Burkhardt, Dirk and Nazemi, Kawa and Kohlhammer, Jörn (2012):
Analytical Semantics Visualization for Discovering Latent Signals in Large Text Collections.
Linköping University Electronic Press, Linköping, In: Proceedings of SIGRAD 2012, In: Linköping Electronic Conference Proceedings; 81, [Conference or Workshop Item]

Abstract

Considering the increasing pressure of competition and high dynamics of markets, the early identification and specific handling of novel developments and trends becomes more and more important for competitive companies. Today, those signals are encoded in large amounts of textual data like competitors' web sites, news articles, scientific publications or blog entries which are freely available in the web. Processing large amounts of textual data is still a tremendous challenge for current business analysts and strategic decision makers. Although current information systems are able to process that amount of data and provide a wide range of information retrieval tools, it is almost impossible to keep track of each thread or opportunity The presented approach combines semantic search and data mining techniques with interactive visualizations for analyzing and identifying weak signals in large text collections. Beside visual summarization tools, it includes an enhanced trend visualization that supports analysts in identifying latent topic-related relations between competitors and their temporal relevance. It includes a graph-based visualization tool for representing relations identified during semantic analysis. The interaction design allows analysts to verify their retrieved hypothesis by exploring the documents that are responsible for the current view.

Item Type: Conference or Workshop Item
Erschienen: 2012
Creators: Stab, Christian and Breyer, Matthias and Burkhardt, Dirk and Nazemi, Kawa and Kohlhammer, Jörn
Title: Analytical Semantics Visualization for Discovering Latent Signals in Large Text Collections
Language: English
Abstract:

Considering the increasing pressure of competition and high dynamics of markets, the early identification and specific handling of novel developments and trends becomes more and more important for competitive companies. Today, those signals are encoded in large amounts of textual data like competitors' web sites, news articles, scientific publications or blog entries which are freely available in the web. Processing large amounts of textual data is still a tremendous challenge for current business analysts and strategic decision makers. Although current information systems are able to process that amount of data and provide a wide range of information retrieval tools, it is almost impossible to keep track of each thread or opportunity The presented approach combines semantic search and data mining techniques with interactive visualizations for analyzing and identifying weak signals in large text collections. Beside visual summarization tools, it includes an enhanced trend visualization that supports analysts in identifying latent topic-related relations between competitors and their temporal relevance. It includes a graph-based visualization tool for representing relations identified during semantic analysis. The interaction design allows analysts to verify their retrieved hypothesis by exploring the documents that are responsible for the current view.

Series Name: Linköping Electronic Conference Proceedings; 81
Publisher: Linköping University Electronic Press, Linköping
Uncontrolled Keywords: Business Field: Visual decision support, Business Field: Digital society, Research Area: Generalized digital documents, Semantics visualization, Trend analysis, Information visualization, Data mining, Business intelligence
Divisions: 20 Department of Computer Science
20 Department of Computer Science > Interactive Graphics Systems
Event Title: Proceedings of SIGRAD 2012
Date Deposited: 12 Nov 2018 11:16
Export:

Optionen (nur für Redakteure)

View Item View Item