White Paper
 

 
 (c) 1998 James Wen.  Permission is hereby granted to copy, reproduce or redistribute in whole or in parts, the paper contained herein provided that any and every such act includes an acknowledgement of the source and a proper reference to the title, author and URL of the paper.  Redistribution of the entire paper  must  include this copyright statement.  All other rights and protection as provided for by the copyright laws apply to the paper, both explicitly and implicitly.
 

 
Surf Visualization: Seeing More than History
James Wen
jcw29@cornell.edu

ABSTRACT 
Users often impose conceptual abstractions to the world wide web, viewing it as a collection of connected sites rather than pages, for instance.  Prior studies of the behavior of web users have found, however, that abstract concepts important in the field of hypermedia appear insignificant when examining surf logs.  This paper proposes that in order to capture such abstractions, existing tools need to be extended to include elements that serve as aides in extracting and interpreting the various features of the explored hyperspace.  The notion of a surf visualization is introduced to describe the retention and display of higher level abstractions built on top of the steady stream of logged pages captured by a history mechanism.  A system built on this concept is described and the implications are presented. 

KEYWORDS 
World wide web, hypertext, URL manager, history mechanism, surf, visualization, graphic history. 

INTRODUCTION 
Getting lost while surfing the world wide web (www or simply, web) is a problem that constantly faces the user browsing the internet.  URL managers and various history mechanisms found in the major browsers are primarily based on the notion of a history list.  But a history list stores references to the previous pages visited in a linear format and, as such, fails to capture the non-linearity of the hypertextual web.  Furthermore, other concepts that arise from the web's hypertextual nature, such as paths and locality, have yet to be applied to the web [4][6][7].  This may, in part, be due to the fact that current technology lacks the necessary features to handle these abstractions - either to extract them naturally or to present them usefully. 

This paper proposes a system that extends the traditional history tool from one which simply lists a user's surf history to one which also handles contextual information regarding the pages visited and the paths that link the various pages.  Since information generated during a surf session is of a streaming (data is available sequentially over time) rather than random access (all data is immediately available) form, an overview and understanding of the surf environment is accumulated by the user in piecewise fashion.  If the opportunity that exists for the user to enrich and annotate the surf world as it evolves is exploited, it can greatly contribute to the user's orientation while surfing the web. 

THE SYSTEM 
At the heart of the proposed system is a graphical history tool.  Little prior work has been done in this area although MosaicG stands out as one notable exception [2]. MosaicG provides a graph view of a history list but lacks any higher level conceptual enhancements. 

Basic Description 
The figure, below, shows a screen dump of the proposed system, dubbed SurfSerf.  As a reflective system, it logs pages visited as the user surfs through the web. 

 
The display is in tree fashion and allows interactive zooming and scrolling.  The nodes, shown as boxes around text names, represent the pages visited.  A node can be in one of three states: normal, selected (highlighted box) or current URL (inverse text).  At any given instance, only one node can be selected and only one node can be the current URL, although a node can be in both states simultaneously.  When the user selects a node by clicking on it, the node's title, name, URL and comments are displayed.  The title and URL are taken from the browser and cannot be modified.  The user is free, however, to modify the name and the comments for additional annotation.  By default, the system creates a name, as best it can, from the given URL. 
Since the web is a fully directed, cyclic graph rather than a well behaved tree, displaying all the links can result in a cluttered and unreadable screen [5].  The solution is to show the primary surf path - i.e. of all the links followed to reach some node A, only the first link taken to A is shown by default.  However, the user has the option to see all the links taken from the selected node or all the links taken to reach the selected node, as well as simply all the links in existence. 
Nodes can be searched by URL, title,  name, or phrases found in the comments.  Unwanted nodes that clutter up the graph can be pruned from the graph through deletion.  Finally, surf sessions can be saved, retrieved and merged with other surf sessions. 

Surf Visualization 
The user-defined mnemonics and graphic view described  offer a core for surf visualization.  On top of this foundation, access to higher level abstractions can be built. 

A locality is defined as a collection of pages that a user surfed within to a higher degree than elsewhere.  While some empirical data based on user logs found little to substantiate the conjectured significance of the notion of localities [6][7], the idea seems so important that users are willing to manually establish them via bookmarks [1].  This discrepancy between studies may be due to the fact that existing systems do not have the sufficient facilities for tracking localities.  With so much to explore on the web, users may find themselves losing track of pages they would otherwise like to revisit.  Using bookmarks to save page references often results in a bookmarking system that is ineffective both because it is overwhelmed and because it mixes noteworthy pages with ones of passing interest.  If, however, pages could easily be marked without requiring the commitment of bookmarks, users might be encouraged to track and return to pages of  interest.Groups of pages so marked can thus mature into localities.  To facilitate this, SurfSerf allows the user to color nodes and to annotate the colors in order to create a contextually rich graph that makes the user's meaningful groupings of related nodes apparent to the user. 

A path between page A and page B is defined as the links taken to reach B from A.  SurfSerf is able to detect and highlight all the paths that brought the surf session from A to B and the user is allowed to scroll through and see all the paths found. 

SurfSerf additionally introduces the notion of a neighborhood in web browsing.  Whereas a path is a series of links taken to get from A to B, the neighborhood of A are all the pages that could be reached from A after traversing n links.  Neighbors of A can be highlighted with each successive generation growing dimmer in intensity.  The notion of a neighborhood allows a user to see all the pages a particular page has taken the session to in the past. 

Finally, SurfSerf subsumes the notion of a history list and combines the lists of different windows into an overall history log.  While some studies have dismissed the importance of the use of multiple windows in a browsing session [6][7], others have embraced its use as fundamental to effective surfing [3].  Either way, the possibility that a surf session may extend over multiple windows should be accounted for by a surf visualizer - the user may, for instance, want to scan the results of a search in new windows so as not to have to backtrack to the original search results.  SurfSerf includes all the windows associated with a browsing session in the same graph, visually tagging pages opened in new windows. 

Applications and Implications 
The functionality of a surf visualizer like SurfSerf suggests many useful applications.  A pruned graph saved to disk can, for instance, serve as a contextual bookmarking system.  A training or presentation tool can be created where the history log is played back to the user simultaneously through the visualizer and browser - in effect providing a graphical tour of how to surf a portion of the web.  The ability to merge surf sessions allows SurfSerf to become a useful utility in large research projects involving multiple surf sessions over time or distributed over many users. 

While SurfSerf can easily find many applications areas, a most interesting implication is one that does not use it as a surf visualizer or history mechanism at all: users have remarked that SurfSerf gives them a "sense of place".  Even when not used for recalling previously visited pages, simply having such a tool around gives the users with a sense of security.  In this way, SurfSerf is very much like a common street map: one can use it to return to a given place but one can also simply carry it around to feel freer when exploring a new place since there is the feeling that getting lost will not be such a great problem with a map in hand. 

This result suggests that current users may not be as exploratory when surfing the net for fear of getting lost.  It may very well be the case that the observation that users venture "rarely more than two links away before returning" [4] can be traced to a fear of getting lost rather than finding that there is nothing worthwhile to explore beyond two links.  SurfSerf's greatest contribution may, in fact, be as a means to encourage greater exploratory surfing on the web. 

CONCLUSION 
This paper has presented a graphical history tool that does more than simply capture a user's history list.  It strives to help the user visualize a surf session by proposing the means to accept and display additional information a user may typically want to associate with the pages visited.  These are features that add context to a history list in an attempt to capture more closely the environment a user envisions when surfing through the web.  The result is a system that not only has immediate applications beyond being a history tool but one that also encourages greater exploration of the web. 

REFERENCES 
[1] Abrams, David and Ron Baecker.  How People Use WWW Bookmarks.  SIGCHI '97    Conference  Proceedings, 1997, ACM Press. 
[2] Ayers, Eric Z. and John T. Stasko.  Using Graphic History in Browsing the World Wide Web.  World  Wide Web Conference Proceedings, 1995. 
[3] Brown, Marc H. and Robert A. Shillner.  DeckScape: An Experimental Web Browser.  World Wide  Web Conference Proceedings, 1995. 
[4] Catledge, Lara D. and James E. Pitkow. Characterizing Browsing Strategies in the World-Wide Web.   World Wide Web Conference Proceedings, 1995. 
[5] Eades, Peter and Roberto Tamassia.  Algorithms for Drawing Graphs: an Annotated Bibliography.   Technical Report CS89-09, Brown University,  1988. 
[6] Tauscher, Linda and Saul Greenberg.  How people revisit web pages: Empirical findings and  implications for the design of history systems.  In. J. Human Computer Studies, in press. 
[7] Tauscher, Linda and Saul Greenberg. Revisitation Patterns in World Wide Web Navigation.  SIGCHI  '97 Conference Proceedings, 1997, ACM Press.