Abstract 1 1 Introduction 2
WebBook™ and Web Forager™
Download 324.2 Kb.
|
- Navigate this page:
- WEBSOM
- WebTOC
- WWW3D
- Appendix B: Research Design
- Hypothesis
- Design
- Reference List
WebBook™ and Web Forager™ 
 
References Card, S.K., Robertson, G.G. and York, W. The WebBook and the Web Forager: an information workspace for the World-Wide Web. Human Factors in Computing Systems: Common Ground: CHI'96 Conference Proceedings, April 13-16, 1996, Vancouver, British Columbia, Canada. WebBook™ is a 3D interactive book of HTML pages. It supports rapid interaction with objects at a higher level of aggregation than pages, i.e. a book is a collection of pages. Web Forager™ is an application that embeds the WebBook™ and other objects in a 3D workspace. Given a collection of web pages, WebBook™ preloads those pages and displays them as a book, one Web page per book page. Page flipping is animated. Links are color encoded and make a distinction between links to pages in the book, another book (red color), and outside the book (blue color). Selecting a link flips to a new page, opens a new book, or spawns a new viewer for external links. Page collections can be automated by any number of mechanisms: following home page links, collecting relative URLs, using hot lists, searching result. Font family and size can be made consistent for all the pages in a book. Books may be viewed in different ways, e.g. using a Document Lens. Bookmarks are represented graphically to correspond to those found in the physical world. Web Forager™ arranges pages hierarchically in terms of usage. "Focus Place," "Immediate Memory," and "Tertiary Place" are the main levels. Focus Place is easy to access and contains the pages that a user has spent a lot of time with. These show up in front view. Immediate memory is depicted as a page or book on the desktop. Objects can be moved around in space, including back and forth in the depth direction. Objects in the back are smaller than objects in front. A Tertiary Place is a place where pages and books are stored, e.g. book case. In normal view, a book case is shown at an angle to reduce presentation space. WEBSOM
Reference: Kohonen, T., Self-organization of very large document collections: State of the art. Niklasson, L., Bodén, M., and Ziemke, T., editors, Proceedings of ICANN98, the 8th International Conference on Artificial Neural Networks, 1998, volume 1, pages 65-74. Springer, London. Lagus, K., Kaski, S., Honkela, T., and Kohonen, T., Browsing digital libraries with the aid of self-organizing maps. Proceedings of the Fifth International World Wide Web Conference WWW5, May 6-10 1996, Paris, France, volume Poster Proceedings, pages 71-79. EPGL. URL: http://websom.hut.fi/websom/ http://websom.hut.fi/websom/comp.ai.neural-nets-new/html/root.html WEBSOM projects documents into meaningful 2D maps for exploration and search. It is based on self-organizing maps (SOM), a general unsupervised learning algorithm for analyzing and visualizing high-dimensional statistical data. The SOM is a nonlinear projection from a high dimensional space onto a low-dimensional grid. A representation model of some subset of data is associated with each grid point. From text documents, a vector of text is used to represent each document. A vector of text is a set of words, excluding stop words, and a value, e.g. weighed word frequency count or inverse document frequency. The dimension of the vector is reduced for feasibility in computing. Originally, WEBSOM use a self-organizing semantic map, semantic SOM. The SOM algorithm is based on the average short context of the words. Words are clustered into an ordered set of words categories. Recently, the randomly projected word category histogram and randomly projected word histogram methods have proven to be good approximations of the corresponding relations between the original document vectors at reasonable computation speed. The interface to WEBSOM provides four levels of viewing the document map; the whole map, the zoom map, the map node, and the document view. WebTOC 
Reference: Nation, D. A., Plaisant, C., Marchionini, G., and Komlodi, A., Visualizing websites using a hierarchical table of contents browser: WebTOC. Human Factors and the Web conferences, June 1997,Colorado. URL: http://www.uswest.com/web-conference/proceedings/nation.html
Reference: Snowdon, D., Fahlen, L., and Stenius, M., WWW3D: A 3D multi-user Web browser. WebNet96 Proceedings Online, October 15-19, 1996, San Francisco, California, USA. URL: http://www.crg.cs.nott.ac.uk/~dns/vr/www3d/webnet96-final.html WWW3D is a 3D WWW browser. The implementation supports a virtual reality system. An HTML page is parsed and the current document and links are presented as small icons with titles on the inside surface of the spheres. Unfolding the current document icon shows its context. When a link is selected, a new sphere is created, which is placed in 3D space based on a "Force Directed Placement" algorithm. Viewed from the outside of the sphere, Web pages are shown as opaque spheres, and links are shown as arrow lines. The colors indicate how long ago a user last visited a document. Appendix B: Research DesignIntroductionThis research examines the effect of integrating navigation tools. While each navigation tool may operate optimally in a certain environment, multiple integrated tools may be useful in a wider range of environments. A document space with a network structure, e.g. hypertext or WWW, is a general class of document spaces. A space that has a list structure or a hierarchical structure is a special case of a network. Two common navigation tools that are frequently used in a network structured document space are: a text viewer with visible anchors connected to links; and a structural overview browser, which is frequently graphical. The text viewer with anchors is the most common mode of navigation in the WWW. In the WWW, the text viewer with anchors is called a “Browser” -- a term historically reserved in the early hypertext literature for a graphical overview. The term Text Viewer with Anchor Links (or simply Viewer) will be used to refer to a WWW like client. The term Graphical Browser (or simply Browser) will be used for a tool that provides a structural overview graphically. A Viewer provides navigation capability as well as document content presentation. Only one document at a time is presented by a viewer. It is similar to navigating in an egocentric view. In contrast, a browser presents a view of the overall structure of a hypertext, an exocentric view. Depending on the size of the document space, a browser may present only a local overview of space. With a scroll bar, other areas can be shown. A browser navigates a document space via active graphical objects. Navigation is a part of a certain task. Navigation for information finding in “first-encounter environments” will be addressed. A “first encounter environment” is a document space within which users find themselves for the first time as they search for information. The extent, organization, and content are not known. This kind of task will become more common especially in the WWW context, which has enormous information stores. Navigation tool performance is measured in many ways including, time to completion of a task, number of nodes visited, number of nodes revisited, etc. In an information-finding task, time to completion may be the major concern, in conjunction with a correctness of the answer. A viewer and a browser are optimized for navigation of different kinds of spaces. In general, the viewer will be more usable in a highly structured document space. For instance, a linear list hypertext may be visited in order, using a viewer. In contrast, a highly a complicated space will be better navigated with a browser. Every node is only one click away in a browser. More abstractly, browsers will be of more use where the user either has or is able to construct a simple mental model of the document space. A viewer provides more information about nodes and links than does a graphical browser because it can show more semantic information about the node in a limited display space. The context around the anchor allows more accurate path selection than is possible in a graphical browser when only the name of a node is visible. The semantic relatedness between information needs and node name and context play a major role in navigation. If there are no hints from the node name in a graphical browser, the advantage of being “one click away” cannot be used. In contrast, in a highly complex environment, with multiple links between nodes and minimal intermediate path information, browsers may offer significant advantages, both in the short run information finding task and in the long run space structure modeling task. Boyle and Teh (1992) show that increasing the number of links, in a preserved hypertext structure, will decrease the average number of nodes visited in an information-finding task. They also find that total time to completion of task and number of errors were not affected by the number of links. Schoon (1997) shows that the different hypertext structures linear, hierarchical, star, and arbitrary, have a significant effect on navigation. The star and hierarchy structures are more navigationally effective than linear. The arbitrary is significantly less effective than others. In conclusion, it is believed that the navigation tool performance will depend on the semantic relatedness of information presented to the information needed and complexity of a document space. The complexity of a document space may be described by a document space size and its structure. However, document structure described in Schoon’s taxonomy is subjective. The following measurements: average distance between nodes and the deviation of distance between nodes are proposed. In space where there is a strongly semantic relation between the question and the content or node name, a user knows where to navigate. In the case of the viewer, the structure of document space will impact navigation performance. For example, a star configuration will allow one-step finding. In contrast, a linear list will require n steps where n is the distance between the original node and the target node. In the case of the browser, one traveling can find the answer’s node. In space where there is a small semantic relation, navigation may be considered as a random walk through the space. Size, average distance between nodes and variation of distance between nodes will cause the difference in performance in navigation. However, the effect will not be equal in using the viewer and the browser. HypothesisThe Null hypothesis of this experiment is: H0: There is no statistically significant difference in the performance of search with integrated navigation tools and individual navigation tools. The working hypothesis is: There are significant differences in performance among different navigation tools in a certain environment. Integrated navigation tools, the browser and the viewer will be more useful in information finding and navigation within complex hypertext spaces than will the viewer or the browser.
Integrated navigation tools will result in faster time to locate information in complex space. Information finding will be more difficult in complex spaces with minimal semantic information in links. The complexity of an information space may be predicted in terms of three measurable attributes -- size, average distance between nodes, and deviation of distance between nodes. The performance of integrated navigation tools will degrade with the simplicity of the space as they become noise contributors rather than information providers.
Other navigation facilities such as the history list, content index, search and history-trail will not be used. The display will be of a fixed size and structure. The complexity of the document space will be varied. The complexity will be measured by three variables: number of nodes (size), average distance between nodes, and variation of distance between nodes. The document space size condition will be small (20 nodes) or large (100 nodes). The small condition will be a subset of a large condition. The average distance between nodes will be short and long. In the short distance condition, the average distance will be 1-2 nodes apart. In the long condition, the average distance will be 10 nodes apart. The variation of distance between nodes will be small and high conditions. In the small condition the variation will be zero. In the high condition, variation will be higher than 10. A complex document space is a space that is large, has long distance between nodes, and high variation of distance between nodes. In contrast, a simple document space is a space that is small, has short distance between nodes, and low variation of distance between nodes. Given an optimal (simple or complex) site, the ability of an individual to identify nodes or paths bearing information is a matter of the semantic relatedness of the node or the anchor/link. The relatedness will be measured in a separate experiment. Ten subjects will be asked to weigh the relatedness of node names, which will ultimately be presented in the browser, on a scale of 1 - 10. The relatedness degree will be variant values of weight. The high variant indicates that the question is not referred to by only some specific node. Ten questions will be prepared with different variant values. The same weighing will be carried out for anchor points. It is anticipated that highly related node names will favor browser use while highly related anchor names will favor anchor use. Four questions will be selected, one which does not favor either navigation tools, one which favors the browser but not the viewer, one which favors the viewer but not the browser, and one which favor both tools. The experiment will be 3 x 2 x 2 x 2 x 4 factorial design: three navigation tools, two document space sizes, two average distances between node lengths, two variations of distance between nodes, and four questions. There are 96 conditions. To minimize subject characteristic variation and number of subjects, each subject will perform the experiment in three conditions, i.e. using different navigation tools. To minimize effect of learning about a document space, three different hypertexts will be used to create equivalent document space. The Latin square will be used to eliminate ordering effect. The task will be to find information in a hypertext. The performance of the navigation tool will be measured by the following: Time to complete task Number of nodes visited Number of nodes re-visited per total number of nodes visited Completion of task and correctness of answer
192 (32 x 6) subjects will be recruited from the University student population. Subjects will be paid. Each subject will perform three conditions using different navigation tools. Each experiment is expected to be complete in 15 minutes. Material Subjects will perform the experiment in a University laboratory. The hypertexts will be stored locally on each machine to reduce delay time. Three hypertexts will be prepared. These three hypertexts will be modified to be small/large size, short/long average distance between nodes, and small/high variation of distance between nodes. One dummy hypertext will be prepared for practice. Procedure Subjects will be randomly assigned to experimental conditions. Subjects will practice using all three navigation tools with the dummy hypertext. Subjects will use each tool to answer a question. Subjects will fill in a questionnaire to obtain demographic information. Data analysis The design is 3 x 2 x 2 x 2 x 4 factorial design. To analyze this data, an analysis of variance (ANOVA) will be applied. Each navigation performance measurement will be treated separately. The semantic relatedness effect in navigation performance will be shown by ANOVA procedure, comparing the mean of each navigation performance of the four questions to determine if a significant exists among them. The effect of complexity of a document space navigation performance will be shown by ANOVA procedure, comparing the mean of each navigation performance of the two sizes, two average distances between nodes and two variations of distance between nodes among the three tools. The interaction effect of each dimension of the complexity will also be assessed. The interaction of semantic relatedness and the complexity of a document space will be shown by ANOVA procedure. The navigation performance of the integrated navigation tool will also be compared with another two navigation tools by the overall mean with t-test. Reference ListDownload 324.2 Kb. Share with your friends:
|
