Visualized Index-Based Search for Digital LibrariesIntroductionMethodsOverview VisualizationDetailed VisualizationExperimentSurvey ResultsConclusion and Future WorkReferencesG. Bebis et al. (Eds.): ISVC 2009, Part I, LNCS 5875, pp. 125–134, 2009. © Springer-Verlag Berlin Heidelberg 2009 Visualized Index-Based Search for Digital Libraries Jon Scott, Beomjin Kim, and Sanyogita Chhabada Department of Computer Science Indiana University-Purdue University Fort Wayne, IN, U.S.A. [email protected], [email protected] Abstract. As the amount of available information continues to expand, tradi-tional text-based searches for digital libraries and similar systems become in-creasingly cumbersome to the user. Selection of the best result calls upon the user to compare and contrast top results; this can involve investigative reading of each, to determine what quality and amount of the desired topic is present in each. This paper presents an alternative search strategy, utilizing visualization to relate detailed content information obtained through indexes. By providing such information in a visual manner, the aim is to reduce the burden of investi-gation placed upon in present systems. 1 Introduction The proliferation of computing technology and the advance of the Internet has greatly changed the way in which people obtain information. In these days, a person can access to the information which is more than we can handle without the limitation of time and location. The amount of available data grows ever rapidly, but the ability for a user to find their desired information has progressed with less vigor. This is particu-larly true of highly concentrated sources of a broad array of information, such as doc-uments on the Internet and Digital Libraries. The current library system provides several attributes associated with books as a response to the users’ inquiry. The search results include book title, author, publica-tion year, ISBN number, thickness, etc. However, there is often a large expectation left on the user’s ability to read through the results. Furthermore, the text-based ap-proach is non-intuitive and inefficient for finding suitable information through com-parison of many possible search results [1]. While ranked search results may assist the user in this endeavor, there is still a reliance on the user investigating the top re-sults individually [2, 3]. This will only become more problematic as the information domain they are applied to continues to grow larger and more complex. Information visualization is an effective tool that can present a large amount of data compactly, but intuitively for easy comparison. By exploiting users’ perceptual cognition, studies have shown that the graphical illustration of data has contributed in improving the users’ understanding and reviewing speeds [2, 4, 5]. Borner and Chen explained that there are three common usage requirements for visual interfaces to Digital Libraries: First, to support the identification of the composition of retrieval result; second, to understand the interrelation of retrieved documents to one another,126 J. Scott, B. Kim, and S. Chhabada and last, to refine a search, to gain an overview of the coverage of a Digital Library (DL) and to facilitate browsing and to visualize user interaction data in relation to available documents in order to evaluate and improve DL usage [6]. When using physical books, people tend to view multiple at once; to better com-pare and review information across multiple sources, and to have a better overall understanding of the domain. In their study, Good et al, identify this to be a major weakness in current DL displays [1]. To address related issues, researchers have con-ducted studies applying visualization techniques for book searches and presenting various forms of search results [7, 8]. The Graphical Interface for Digital Libraries (GRIDL) is a system that displays a hierarchical cluster of the relevant data to a query on two-dimensional display [9]. This system uses a two-dimensional coordinate, the axes of which are selectable from a variety of different attributes. Results were displayed within each cell as a collection of different size icons, color coded by document type. Marks and his colleagues pre-sent a similar approach, based on scatter plots, known as ActiveGraph [10]. Because this approach results in much more node clustering and overlap, a logarithmic trans-formation is provided, along with the ability to filter out user specified documents. ActiveGraph also provides the ability to specify shape, color, and size of nodes repre-senting documents. By allowing users to manipulate the manner in which data is displayed, these visualizations provided a strong ability to reveal patterns within the data that may not typically be apparent. These studies mainly focused on aiding the user in comparing the search results ef-fectively by presenting book properties through various visual attributes; but they don’t express in detail the amount of content related to user interest. Lin proposed a graphical table of contents (GTOC) that showed the dimension of items in the table of contents based on Kohonen’s self organizing feature map algorithm [11]. The paper introduces how documents can be organized and then visualized to allow the user easy access of underlying contents. The GTOC prototype describes various interac-tive tools to assist the user exploring document contents and analyzing relationships among terms in the table of contents. The main goal of the research presented in this paper is the development of visuali-zation techniques that will make the user’s book search effective by exploiting the book index. 2 Methods The Visualized Index-Based Search (VIBS) system utilizes an Overview + Detail approach for presenting book search results. This is a visualization technique that uses multiple images to display the entire data space, as well as show an up-close, detailed view of the data [12]. Similar to traditional library searches, the overview will present outline of the book search results through graphical illustration. The user interactively selects a subset of visualized icons that will allow them to execute content level ex-ploration. When a user provides search terms of interests, the Detail view presents a rich visualization of the assets of the given query in a book index with other related information. The resulting