HyperSpace: Web Browsing with Visualisation

Andrew Wood, School of Computer Science, University of Birmingham


A.M.Wood@cs.bham.ac.uk

http://www.cs.bham.ac.uk/~amw/

Nick Drew, School of Computer Science, University of Birmingham


N.S.Drew@cs.bham.ac.uk

http://www.cs.bham.ac.uk/~nsd/

Russell Beale, School of Computer Science, University of Birmingham


R.Beale@cs.bham.ac.uk

http://www.cs.bham.ac.uk/~rxb/

Bob Hendley, School of Computer Science, University of Birmingham


R.J.Hendley@cs.bham.ac.uk

http://www.cs.bham.ac.uk/~rjh/

School of Computer Science
The University of Birmingham
Edgbaston, Birmingham B15 2TT
United Kingdom

Keywords:: World-Wide Web Visualisation, Browsing, Mapping, Virtual Reality

Introduction

The W3 is a vast collection of geographically distributed, essentially unorganised information, and whilst there is likely to be the answer to your question out there, it can be impossible to find it. Worse, once you have discovered relevant information, it is unclear where to go next to find further information, and even when you choose an interesting path, rediscovering one that you noticed earlier is difficult. Users become lost in the maze of hypertext links, and need support in navigating the web.

HyperSpace

HyperSpace is a prototype World-Wide Web visualiser that can be used to display the organisation of areas of the web. It structures the information not according to geographical location, but according to a user-defined structure, which means that related topics are displayed adjacent to each other, and unrelated topics are spatially separated.

Each page on the web is represented as a sphere, and links from one page to another are represented as links between the spheres. These spheres and links are placed into a 3-d virtual reality system, initially randomly. The chaotic and unstructured mesh of nodes and links is then allowed to self-organise according to some imposed physics within the reality. Nodes repel each other, whilst links provide an attractive force. Thus, unrelated areas that do not have links between them are pushed apart, whilst highly-interrelated work is pulled together and clustered in the same region of space.

Characteristic structures form that serve as landmarks to aid and guide the navigation process. From any page that has been visited, HyperSpace provides a view of all the pages that are linked to that node; when you encounter a page for the first time you can immediately see all the other pages that lead away from it; the viewer also shows all the incoming links from any of the other pages that it knows about.

The System

Our system collects URLs using an adapted browser that passes the current URL to an external program; each time the user moves to a new page the new URL is passed out. This is picked up by a separate process that fetches the page again (from a local cache), parses the content and extracts all the links and the current page title. All the data is collated together, to be used by the HyperSpace viewer. This assigns each URL a sphere whose size is dependent on the number of links and a random spatial position. The system can then self-organise (collapse) into the organic structures shown below.

Some Results

A ray-traced image of the initial random structure reflected in one of the page nodes, with an MPEG movie showing the web structure moving from its starting chaotic structure to its final organisation.

A ray-traced image of the structure formed by a fully explored set of home pages (notice how pages that have not yet been explored tend to lie on the outskirts of the structure).

A ray-traced image of the structure of several index pages on similar topics.

A ray-traced image of the structure of over 750 pages. In the foreground is a highly cross-referenced set of manual pages, further in the distance several large index pages can be seen.

A ray-traced image of the structure formed from an actually browsing session (notice the `constellations' formed by completely separate unconnected pages).

A screen-shot of the HyperSpace system in use, with the viewer window, Mosaic browser, and various control panels.

A screen-shot of the viewer window with the structure only showing the links from the selected page [1], and also the names of pages (unexplored pages simply show their URL).

Future Developments

The current system allows users to view the `map' of their progress through the web at any stage during their exploration, but on each occasion the view has to be reloaded and collapsed (with a significant time overhead). We are currently developing a version of the viewer that will add the pages as the user accesses them, and will re-organise itself automatically; allowing the user to see immediately where they are. The viewer and the browser will also be synchronised, enabling the user to access a page in the browser simply by clicking on its sphere in the viewer.

We are also working towards adding extra customisation features to the viewer, allowing the user to specify the size, shape, colour and names of the nodes. Changing each of these parameters, even subtly, can affect which features the visualisation highlights. For example, currently the size of the sphere is related to the number of links that are present on the page, in order to allow enough space to render the objects. It may be better to collect usage or response statistics on that node and set the sphere size accordingly, thus cueing the user as to popular or rapidly accessible pages. Other planned improvements and enhancements are detailed in [HSV95].

We believe that allowing users to visualise the web in this way will help them to orient themselves within its information landscape - allowing them to make more effective use of the many resources that the web provides.

[1]: The viewer display without links has been nicknamed affectionately `cauliflower space'.

[HSV95]: Wood, A, Beale, R, Drew, N & Hendley, R, "HyperSpace: A World-Wide Web Visualiser and its implications for Collaborative Browsing and Software Agents", submitted to HCI'95, UK.