Navigation in Huge Information Hierarchies - CiteSeerX

2 downloads 0 Views 204KB Size Report
There is a need for meta-content standards; there is [MARC], but it is not .... [Beau96] Beaudoin L., Parant M.-A., Vroomen L., Cheops: A Compact Explorer For.
Navigation in Huge Information Hierarchies Bénédicte DESCLEFS [email protected] http://www.lip6.fr/rp/~desclefs Université Pierre et Marie Curie Laboratoire d’Informatique de Paris 6 (LIP6)

Michel Soto [email protected] http://www.lip6.fr/rp/~ms Université Pierre et Marie Curie Laboratoire d’Informatique de Paris 6 (LIP6)

Abstract Huge information hierarchies are very difficult to represent and users often find it difficult to retrieve relevant information from such a mesh of nodes. The Web is a complex hierarchy: users need to understand the whole structure in order not to feel lost, and they also need to be able to find the nodes that are likely to be relevant to them. Most representations become cluttered very quickly. In this paper, after reviewing visualization techniques for large hierarchies, we propose enhanced tools for information aggregation. This approach allows a significant reduction of the number of nodes to display, so that the user management of data is easier. These techniques can be used for information that is retrieved from the Web, as well as from any other hierarchy.

Keywords: Hierarchies representation, 3D visualization, navigation, information retrieval and aggregation

Introduction This work started a few months ago. It is partially funded by an industrial contract and aims at investigating new 3D visualization tools and paradigms for large information hierarchies. The starting point of this work was initially to represent communications between physical ports of network equipment. We decided to design a generic tool to be used with any type of hierarchy, for example the World-Wide Web. The volume and lack of structure of the data contained in the Web make navigation and information retrieval extremely difficult. When a user is connected to an Internet site, he sometimes has problems understanding the hierarchy he is exploring. The use of hyperlinks

1/15

may confuse him; he does not manage to explore the site efficiently and is thus unable to find the pieces of information that are interesting for him. The aim of this research is to assist the user in information retrieval from large hierarchies. The Web is an example of such a hierarchy, but our aim is to apply our results to any structure that needs to be explored in order to find information (for example, some information retrieval tools organize their results in trees, through which the user can navigate, instead of organizing them in lists). When the number of nodes in a hierarchy increases, the visualization becomes cluttered. Traditional 2D visualization tools are not very useful when there are too many nodes to display, because nodes are too small to be distinguished easily. It becomes difficult to retrieve information from one node in particular, as it is “lost” within the global structure. It is of course necessary to be able to focus on specific nodes, but the user should also be allowed to have the global structure in mind. We need: - To reduce the user's feeling of being lost, by maintaining the global context of the hierarchy all the time, - To assist him in his information retrieval, i.e. in finding the nodes that are relevant to him. With current technologies, it is possible to visualize the whole context, but finding a specific node can be really difficult. This paper is structured as follows: section I and II are states of the art in data representation and tree simplification techniques. Section III presents our approach that can be shortly described as an automatic pruning of the tree with information retrieval and aggregation techniques.

I. Visualization techniques and tools In information visualization, both representation and navigation are important. First, the way nodes are displayed is essential, as this is a critical issue when the size of the hierarchy increases: representations deal with this problem. The techniques and tools for data representation provide the user with an overview of the hierarchy and enable him to visualize a specific node within the whole context. However, one must keep in mind that visualizations are used for information retrieval, and not only for understanding the global structure of the hierarchy. The interface should allow the user to manipulate data, in order to organize it the ways he wants: this is the concept of dynamic visualization. While navigating within a hierarchy, one should be able to find a specific node and to visualize it within the global environment. It is therefore important to get the user completely involved in the visualization. Fundamental factors for a good visualization interface are ([Kurn94]): • An overview of the structure for a global understanding of the structure and of the relationships within the hierarchy, • The ability to zoom and to select some nodes, • Dynamic requests in order to filter data in real time. Moreover, the “ideal” visualization tool should be easy to use and adaptable to standard PCs. We are now going to study how existing techniques answer these needs, and how they may be enhanced. In part I.1, we will focus on representation tools; then we will consider dynamic techniques improving navigation (Part I.2).

2/15

I.1. Methods for data representation: overview of the hierarchy and ability to focus In order to help the user keep the global context in mind, we need to be able to display the whole hierarchy. This visualization should enable him to understand the structure and the main relationships within the tree, as well as allowing him to focus on specific parts of the structure. Let us review a few metaphors used for hierarchical information visualization: -

Treemaps ([Asah95]) represent hierarchical information via a 2D rectangular map, providing compact visual representations of complex data spaces through both area and color. This representation uses the full screen very efficiently.

Example of Treemap

-

Information landscapes ([Andr98]) display information in 3D. [FSN95] is an information landscape used for file systems navigation: it lays out the directories in a hierarchy with each directory represented by a pedestal. The height of the pedestal is proportional to the size of the files in the directory. The directories are connected by wires, on which it is possible to travel. On top of each directory are boxes representing individual files. The heights of the boxes represent the size of the file, while the color represents the age.

Example of file system visualization with FSN 3/15

-

Information pyramids ([Andr98]) are another metaphor in which a plateau represents the top of the hierarchy while other, smaller plateaus arranged on top of it represent its subtrees; pyramids thus grow upwards as the hierarchy descends.

The two latter techniques are very useful for small hierarchies visualizations, as the interfaces they provide are intuitive. Unfortunately, as they are quickly cluttered and may not be fitted to large hierarchies.

Focus + context techniques: These techniques make it possible to focus on a detail while including it in the whole hierarchy’s context: - Fisheye Visualizations ([Sark93]): these visualizations show the center of interest in a large scale and with great detail, while areas further from the center are successively smaller and less detailed. Examples of tools using this technique are Perspective Wall and Magic Lens.

Example of Perspective Wall -

Hyperbolic visualizations ([Munz95]): with hyperbolic geometry, it is possible to display trees that grow exponentially, while in Euclidian spaces representations are rapidly cluttered. The idea is to lay out the hierarchy in a uniform way on a hyperbolic plane and map this plane onto a circular display. This representation is less esthetic than information landscapes, but many nodes can be displayed.

Example of hyperbolic tree

-

Cone trees ([Xerox]) and similar techniques are 3D interactive visualizations of hierarchically structured information. Each subtree is associated to a cone; the vertex at the root of the subtree is placed at the apex of the cone and its children are arranged

4/15

around the base of the cone. According to [Tver93], the problem with the first version is that it is difficult to find a specific node within the whole structure. For large hierarchies, a good distribution of nodes is not enough to provide an efficient visualization. SPAVIS implementation of the cityscape metaphor ([Kesk97]) enhances cone trees by giving redundant information about hierarchical information. The importance of nodes is symbolized not only by their position within the structure, but also by their size: the more important the node is, the bigger it appears in the visualization. One can also use other visual cues, such as nodes’ shape, orientation and texture. [Carr95] also enhances cone trees by using fisheye views, colors, shapes and text. Using shapes rather than texture is interesting as different shapes are easy to distinguish, even with a poor resolution. Text can be added to give more information about a node (for instance the title of the Web page). The use of different levels of detail improves the rendering and reduces visual cluttering.

Example of Cone Tree Focus + context techniques fragment information by displaying a part in detail while maintaining the global context. These systems all use the concept of DOI (Degree Of Interest), which is a problematic issue for large hierarchies. It uses the distance between the focus and the a priori importance of information. When the focus changes, the position of every node must be recalculated, which degrades the performance. These methods are not fitted to large hierarchies: cone trees, for instance, cannot display more than 5000 nodes. -

With the Cheops system ([Beau96]), no DOI function is used, so there is no need to recalculate everything when the focus changes. The Cheops method is based on a compressed visualization of a hierarchical data set, using triangles as visual components. It reuses visual components through the alternate deployment of branches. Visual objects in the last level of the tree represent more than one logical node. These “overloaded” objects become unambiguous when a parent node is selected: when a component is chosen, the hierarchy associated with this node is displayed. This tool achieves an effective compression of the hierarchy; yet the representation is in 2D.

A binary hierarchy

5/15

Basic tessellation of triangles with Cheops

We have reviewed a few general visualization methods for large hierarchies; the following ones are specific to the Web: -

In [Bray96], a metaphor for Web site visualization is proposed. Web sites are characterized by the following parameters: the number of pages they contain (size), the number of sites that points to them (visibility) and the number of links to other sites they provide (luminosity). They are represented as cylinders crowned with globes. The diameter symbolizes the size of the Web site, the height its visibility, and the size of the globe its luminosity. Colors are used to distinguish different domain extensions (red for .gov, green for .edu, blue for .com, etc.). Related sites are physically close to each other in the visualization. This metaphor is very interesting, but it is especially fitted to Web site visualizations, whereas cone trees may be used for any kind of hierarchy.

Example of Web visualization [Bray96] -

Another means of helping the user not to feel lost is to provide him with a history of his navigation. In WWW3D ([Snow96]), a 3D structure is used to represent visited Web pages. Spheres symbolize documents and the content appears as 3D icons inside the sphere. Several levels of detail are used in order to enhance performances. Force Direct Placement algorithm is applied to arrange spheres in the 3D space. Related documents are linked by arrows and colors show the time elapsed since the last visit. WWW can support multiple users, but its drawback is that it does not to show the correlation between documents’ content.

6/15

Example of Web documents visualization with WWW3D -

WebBook ([Card96])arranges pages in books, according to different filters. This technique is efficient to retrieve information but it does not clearly show relationships between books (i.e. the Web’s structure is not emphasized).

Example of WebBook -

As we have seen, most Web browsers only display the document being visualized at a specific moment without indicating the pages’ structure or the way they are related. WebPath ([Fréc98]) makes up for these tools’ limitations; it gives a structure to the network formed by visited documents. This application is based on Virtual Reality; it is intended to be used with traditional browsers to give a real-time visualization of their history. Documents appear as cubes with the title on the top. Cubes are more interesting than spheres for the use of texture as texture is less distorted on cubes. Levels of detail are used to reduce the environment’s visual complexity: when the user is far from the node, the title becomes a polygon, the icon disappears, etc. When a new page is visited, a cube is created, as well as a link. This link’s color indicates if the document comes from the same site as the preceding one. The position on the vertical axis corresponds to the moment the page was downloaded, so that the last visited document always appears on the top. The metrics for the two other axes can be chosen freely: it may be the downloading time, the

7/15

number of images, of links, the document’s size, the last modification date, the location of the server, etc. Then, it becomes possible to retrieve a document with only few indications, such as for instance: document located in Asia, document without images, etc.

Example of Web site visualization with WebPath

I.2. User involvement for information retrieval: data manipulation As we said before, visualizations should be dynamic so that the user can find the information he is looking for. This can be done in cone trees with [Carr95], thanks to manual and automatic animation. The user can rotate the structure to put nodes of interest in front of him. Automatic animation may consist in successive zooming, so as to progressively come closer to a set of nodes. Intelligent filtering also makes information retrieval easier by pruning the tree interactively (thanks to user’s requests). Animation and filtering help the user select interesting areas within the tree. SDM (Selective Dynamic Manipulation) is another technique for interaction with data ([Chua95]): • Selective: the user has a high level of control in objects selection, • Dynamic: real-time interaction, • Manipulation: users can move objects and change their appearance. SDM can emphasize specific objects by coloring them differently, enlarging them or bringing them to the foreground, without losing the global visualization. SDM also enables the user to view distinct elements with different scales. Finally, users can organize data the way they want by building their own classification structure. It is important for the user to be able to manipulate nodes and subtrees. The first step is to let the user interact with the initial hierarchy. The user can delete nodes that do not interest him; the hierarchy thus becomes simpler, as only interesting nodes are kept. The second step is to do this simplification automatically, according to the user's profile and points of interest. The visualization tool can automatically prune the tree and let only relevant nodes (or subtrees) appear on the representation. We will consider this in Part II.

8/15

We have reviewed several visualization methods; some of them are suitable for any hierarchy, others are specific to the Web. As far as the Web is concerned, the tool proposed in [Bray96] is particularly appropriate. Among more general techniques, the compression proposed in [Beau96] is very interesting. Moreover, [Xerox], enhanced with visual cues and animation, are both efficient and easy to use. In order to be able to use them, we need to reduce the number of nodes displayed.

II. Simplification of the initial tree It may therefore be useful to let the user interact with a synthesized structure: instead of displaying the initial (and huge) structure, the user is provided with a "simpler" visualization, and he can “unfold” the parts that seem relevant to him. The different methods we can use to achieve this goal are: • Dynamic requests and intelligent filtering according to the user’s preferences, • Selection of “important” nodes, • Aggregation of information in these nodes.

II.1. Dynamic requests / intelligent filtering There are two types of information retrieval mechanisms: content-based and structure-based mechanisms. With the former, the degree of interest of a page depends on the occurrence of some keywords. Content-based research criteria can be the size of a page, the downloading time, the number of links, of images, etc. The MedExplore project ([Naue97]) takes advantage of both the simplicity of thematic access (using domain-specific knowledge) and the efficiency of keyword access (using search tools): navigation graphs are created to assist the user in his information retrieval on the Internet. Two clustering level are used, according to the general nature (or specificity) of documents in order to provide two different types of graphs: • A general graph, for non-specialist users, • A specific graph, for specialist users. These graphs are a basis that the user can use directly by visiting referenced pages - this is a thematic access to information; he may also do a more precise request with a search tool - this is a keyword access. Artificial intelligence may be used to assist Internet users during their navigation ([Jacq97]). The pages the user prefers are characterized by different criteria; the information retrieval tool then looks for other pages satisfying these criteria. Some systems help the user by giving him links to interesting pages. This is an interesting idea, but the degree of interest of a page is often based only on the observation of words contained in the page, which is not enough. Inductive logic allows us to express relationships between pages; in the INDWEB system ([Jacq97]), each web page is defined by some features, such as its name, its URL, its domain, keywords, number of images, complexity, access number, users’ name, language, etc. Any new characteristic may be added. Users groups are defined, according to a topic or an activity (e.g. cinema, education, reading). With these groups, a community can benefit from a

9/15

member’s discovery. INDWEB’s main advantage is that it can “learn” the user’s preferences from many different features and not only from keyword occurrence.

II.1.2. Selection of important nodes and information aggregation Another solution to simplify representations is to show the context of nodes of the Web with respect only to the most important nodes - “landmark nodes” - of the hierarchy ([Mukh95]) . Criteria must be defined to measure the importance of an object. In [Mukh95], they are based only on the document’s structure. Examples of such criteria are the number of pages pointing to the document, the number of links from this page or the number of images. Information retrieval from the Web is difficult because data is not organized. Search tools do not have a good indexation of the documents they handle and often return irrelevant documents. Let us review a few tools using indexing systems: •

HotSauce ([Kier99]) is an information retrieval tool; its main interest is the data file format: the Meta-Content Format ([DEAT96]). As a Web browser transforms HTML instructions into rich Web pages, HotSauce displays MCF files. Meta-content consists of information about content (for instance, an email header is a meta-content, as it is an information about the message but not the message itself). MCF files are text-based descriptions of a site’s organization. Meta-content about Web pages can be best given by authors themselves. This meta-content may consist of keywords describing the content of the pages. With HTML 3.2, adding meta-content information is possible with the “META” tag. There is a need for meta-content standards; there is [MARC], but it is not open. If each Web site generated an MCF description of itself, one could explore any site and find interesting information, without having to use a site map or any other tool. MCF could make the Internet easier to manage. MCF attaches properties to objects: a Web page may have a property giving its size, another its URL, another its author, etc. The problem with this approach is that there may be too much meta-information; this data should be “summarized”.

Example of Web site visualization with HotSauce

10/15



With Document Explorer ([Fowl96]), the visual navigation environment is based on the semantic content rather than on links’ structure: the reason for this choice is that documents may be “semantically closed” without being linked with a hyperlink. Document Explorer uses traditional indexation and content extraction tools (such as Harvest in [Bowm94]), as well as new abstraction techniques based on graphs and clustering. Harvest indexes documents with keywords. Document Explorer works with a list of keywords, in order to find associations between documents, using a co-occurrence metric. It uses Pathfinder networks ([Dear90]) as association networks. These networks are based on the minimum cost principle; there may also be a maximum link number criterion, which makes it possible to visualize more or less detailed graphs.



WAVE (Web Analysis and Visualization Environment, in [Kent94]) is a 3D interface for visualization and navigation on the Web. It uses conceptual analysis to cluster objects. Autonomous agents can be used to retrieve and analyze information from the Web. Automatic documents analysis can be divided into three phases: • Raw data acquisition, • Analysis and automatic clustering, • Interactive visualization and navigation. There are four ways of gathering information: a “spider” (recursive search, requiring a lot of bandwidth), an index (using Web pages’ features), hand-built catalogs and software agents (Web wanderers) that can cross the border between two computers. [KENT94] uses an index for local information and a simple spider to access other databases. Its aim is to arrange Web documents in conceptual classes showing differences and similarities necessary for a good understanding.



The most common indexing mechanisms we described all have drawbacks: keyword indices associate the semantic meaning with syntactic content, which is not always relevant because of homonyms. Hand-built catalogs require a lot of time and must be updated regularly. Finally, web wanderers generally gather only the limited semantic information inferable from existing HTLM tags. It is thus worth considering another approach. Information retrieval from Web pages may be enhanced by ontology-based knowledge about page contents ([Luke96]). Ontologies are an “official” set of attributes and relationships about a concept: they offer a unified structure for sharing knowledge on the World-Wide-Web. SHOE (a set of Simple HTML Ontology Extensions) gives HTML authors the ability to embed knowledge directly into HTML pages, making it simpler for user-agents and robots to retrieve and store this knowledge. HTML documents are annotated using one or more ontologies; HTML authors may use existing ontologies from common libraries on the Web or define new ones with SHOE.

III. Our proposals We have reviewed a few hierarchy visualization techniques and tools. Cone trees are very interesting, and may be enhanced by animation techniques such as Selective Direct Manipulation. However, they are limited by the number of nodes they can display. Our research thus aims at reducing the number of nodes to display. Our tool is not developed yet, but the remainder of this paper describes its specifications.

11/15

We need to select nodes (as proposed in [Mukh95]), so as to display only the ones that are relevant to the user. The representation will therefore be less cluttered. If the hierarchy is new to the user, an automatic aggregated tree is displayed, in which nodes are selected according to general criteria. The structural ones proposed in [Mukh95] are relevant, but we need to add contextual features, such as the number of visitors of the page. After this introduction to the structure, the user may wish to define his own criteria of interest ; in this case, the tree is aggregated according to these criteria. They can be structural (number of links), semantic (keywords about content) or contextual (author, number of visitors). Here is an example of a simple hierarchy, representing a personal Web page : Jean Dupont’s Homepage Welcome Page

Work

Leasures

Laboratory

Team Members

VR

Personal Research (VR)

Themes

QoS

CV

Projects

Sport

Music

Publications

Multicast

NB) This is a 2D diagram aiming at explaining our method’s rudiments. However, the final visualization will be a cone tree. If the user is interested in Virtual Reality, he can formulate his request with the keyword « VR ». Relevant nodes – called “primary nodes” - are selected : they will automatically belong to the aggregated tree. Some nodes need be displayed even if they were not selected at the beginning, because some of their children are primary nodes. For example, the « Work » node is displayed because it is the father of the primary node: « Personal Research ». Nodes like “Work” are called « secondary nodes ». Aggregated trees are thus composed of primary and secondary nodes. The other nodes are not displayed; they are « discarded nodes ». The following figure shows the aggregated tree obtained from the initial structure after a user request; primary nodes are represented as yellow boxes, whereas secondary nodes are white. If nothing more than primary and secondary nodes was displayed, the user would not be able to retrieve information about discarded nodes; this is a problem, because the user may want to see the page about music, even if his first interest is Virtual Reality. We therefore think it is important to provide information about discarded nodes. Any primary or secondary node contains information about its discarded children, if it has some. This information appears in blue. For example, the aggregated tree shows that the “Welcome Page” node is linked to “Leasures”. “Welcome Page” becomes a complex – or aggregated – node: it contains data about other nodes. It appears as a shadowed box.

12/15

Aggregated tree obtained with the request: "VR" Welcome Page Leasures

Work CV

LaboratoryQoS Team Members

Themes QoS Multicast

Personal Research (VR)

Projects

Publications

VR

The user is free to expand the branches that seem relevant to him. Complex nodes contain data about the discarded nodes below them in the hierarchy; it is a means of having information about them even if they are not displayed. If a complex node seems interesting, the user can “expand” it and visualize its children. The first step will be provide simple information about discarded nodes, such as their title. In the future, we plan to add more precise data about nodes’ content and structure.

Conclusion and future work After reviewing visualization methods, we concluded that cone trees were interesting, as they use 3D and can be enhanced by animation techniques. Since cone trees cannot efficiently display more than a few thousand nodes, we studied means of reducing the number of nodes in the hierarchy. The specifications of the tool we will develop were described; a method was proposed for selecting nodes and for aggregating data about discarded nodes, in order not to lose information. Our work can therefore be divided into two phases: • First, information about the “disappearing” subtree must be gathered (this refers to the issue of indexation of Web pages). Among the indexing tools we have studied, ontology-based knowledge retrieval seems particularly promising. • Second, this piece of information must be organized and aggregated in order to be inserted in complex nodes. Then the tool will be tested and evaluated: we will study the impact of our aggregated tree on information retrieval.

13/15

References [Andr98] Andrews K., Visualizing Rich, Structured Hypermedia, IEEE Computer Graphics and Applications, July/Aug. 1998, pp 40-42. [Asah95] Asahi T., Turo D., Shneiderman B., Visual decision-making: Using treemaps for the Analytic Hierarchy Process, Proceedings of CHI’95 Conference, May 7-11, 1995, Denver, Colorado, USA: ACM. http://www.acm.org/sigchi/chi95/Electronic/documnts/videos/ta_bdy.html [Beau96] Beaudoin L., Parant M.-A., Vroomen L., Cheops: A Compact Explorer For Complex Hierarchies, Proceedings of the conference on Visualization '96, Oct. 27-Nov. 1, 1996, San Francisco, CA, USA: ACM. http://www.crim.ca/~vroomen/writing/technical/cheops/cheops.html [Bowm94] Bowman C. M. et al., The Harvest information and discovery system, Proceedings of the Second International WWW Conference, WWW’94, 763-771, May 25-27, 1994, CERN, Geneva. [Bray96] Bray T., Measuring the Web, Proceedings of the Fifth International World Wide Web Conference, May 6-10, 1996, Paris, France. http://www5conf.inria.fr/fich_html/papers/P9/Overview.html [Card96] Card S. K., Robertson G. G., York W., The WebBook and the Web Forager: an information workspace for the Wolrd-Wide Web, Proceedings of CHI’96 Conference, Apr. 13-16, 1996, Vancouver, British Columbia, Canada. [Carr95] Carriere J., Kazman R., Interacting with Huge Hierarchies: Beyond Cone Trees, Proceedings of the IEEE Symposium on Information Visualization, Oct. 1995, Atlanta, Georgia, USA, pp. 74-81. http://www.cgl.uwaterloo.ca/~rnkazman/HCI-papers.html [Chua95] Chuah M. C., Roth S. F., Mattis J., Kolojejchick J., SDM: Selective Dynamic Manipulation of Visualizations, Proceedings of UIST’95 Conference, Nov. 15-17, 1995, Pittsbugh, Pennsylvania, USA. [Dear90] Dearholt D. W., Schvaneveldt R. W., Properties of Pathfinder networks, In R. W. Schvaneveldt (Ed.), Pathfinder associative networks: Studies in knowledge organization, 165178., 1990, Norwood, NJ: Ablex. [Deat96] Deatherage M., HotSauce and Meta-Content Format, TidBITS Magazine #355, Nov. 25, 1996: TidBITS Electronic Publishing. http://www.tidbits.com/tb-issues/TidBITS-355.html [Fréc98] Frécon E., Smith G., WebPath – A Three Dimensional Web History, Submitted to the 1998 IEEE Symposium on Information Visualization (InfoVis ’98), Oct. 19-20, 1998, NC, USA. http://www.comp.lancs.ac.uk/computing/users/gbs/webpath/ [Fowl96] Fowler R. H., Kumar A., Williams J. L., Visualizing and Browsing WWW Semantic Content, CETAC’96, 1996. http://www.cs.panam.edu/info_vis/doc_explorer/cetac96.html [FSN95] Silicon Graphics, 3D File System Navigator for IRIX 4.0+, 1995 http://www.sgi.com/Fun.free_cool_sw_01.html [Jacq97] Jacquenet F., Brenot P., Apprentissage des préférences utilisateurs pour l’aide à la navigation sur le Web, Proceedings of H2PTM’97 Conference, Sept. 25-26, 1997, Paris, France. [Kent94] Kent R. E., Neuss C., Creating a Web Analysis and Visualization Environment, Proceedings of the Second International WWW Conference, Oct. 17-21, 1994, Chicago, Illinois, USA. http://www.ncsa.uiuc.edu/SDG/IT94/Proceedings/Autools/kent/kent.html

14/15

[Kesk97] Keskin C., Vogelmann V., Effective visualization of hierarchical graphs with the cityscape metaphor, Proceedings of the ACM Conference on Information and Knowledge Management (CIKM’97), Nov. 13-14, 1997, Las Vegas, Nevada, USA. [Kier99] McKiernan G., The Big Picture(sm): Visual Browsing in Web and non-Web Databases, LibraryLand: Internet Librarianship: Organizing and Cataloging the Internet, March 5, 1999. http://www.public.iastate.edu/~CYBERSTACKS/BigPic.htm [Kurn94] Kurnar H, Plaisant C., Teittinen M., Schneiderman B., Visual Information for Network Configuration, University of Maryland CS technical reports, June 1994, Maryland, USA. http://www.cs.umd.edu/TRs/authors/Marko_Teittinen.html [Luke96] Luke S., Spector L., Rager D., Ontology-Based Knowledge Discovery on the World-Wide-Web, Proceedings of the Workshop on Internet-based Information Systems, AAAI-96, 1996, Portland, Oregon, USA. [Marc] MARC Standards, Library of Congress, Network Development and MARC Standards Office http://lcweb.loc.gov/marc/readings.html [Mukh95] Mukherjea S., Foley J. D., Showing the Context of Nodes in the World-Wide Web, Proceedings of CHI’95 Conference, May 7-11, 1995, Denver, Colorado, USA: ACM. http://www.acm.org/sigchi/chi95/Electronic/documnts/shortppr/sm2bdy.htm [Munz95] Munzner T., Burchard P., Visualizing the Structure of the World Wide Web in 3D Hyperbolic Space, special issue of Computer Graphics, ACM SIGGRAPH, 1995, pp. 33-38, New York, USA http://www.geom.umn.edu/docs/research/webviz/webviz/node1.html [Naue97] Nauer E., Lamirel J.-C., Environnement d’investigation sur WWW : assistance à l’utilisateur par des connaissances fédérées, Proceedings of H2PTM’97 Conference, Sept. 25-26, 1997, Paris, France. [Sark93] Sarkar M., Brown M. H., Graphical Sark93 Views, Brown University CS Department Technical Reports: CS-93-40, March 23, 1993, USA. http://www.cs.brown.edu/publications/techreports/reports/CS-93-40.html [SNOW96] Snowdon D., WWW3D: A 3D Multi-User Web Browser, Proceedings of WebNet96 Conference, Oct. 31-Nov. 5, Toronto, Canada: AACE. http://www.crg.cs.nott.ac.uk/~dns/vr/www3d/webnet96-final.html [Tver93] Tversky O. J., Snibbe S. S., Zelenick R., Cone Trees in the UGA Graphics Systems: Suggestions of a more Robust Visualization Tool, Brown University CS Dept Technical Reports: CS-93-07, 1993. http://www.cs.brown.edu/publications/techreports/reports/CS-93-07.html [Xerox] XEROX Information Visualizer http://www.birkhauser.com/hypermedia/cyb59.html

15/15

Suggest Documents