WebView: A Graphical Aid for Revisiting Web Pages
`
`Andy Cockburn1, Saul Greenberg2, Bruce McKenzie1,
`Michael Jasonsmith1 and Shaun Kaasten2
`
`1Department of Computer Science, University of Canterbury, Christchurch, New Zealand.
`{andy, bruce, mpj17}@cosc.canterbury.ac.nz
`
`2Department of Computer Science, University of Calgary, Calgary, Alberta, Canada.
`{saul,kaasten}@cpsc.ucalgary.ca
`
`Abstract
`Current commercial web browsers such as Netscape Navigator and Microsoft Internet Explorer provide a wide and
`diverse range of utilities, such as history lists and bookmarks, that support revisiting previously seen pages on the
`web. Yet previous research indicates that these utilities are largely unused. In this paper, we present an alternative
`utility called WebView; a prototype designed to improve the efficiency and usability of page revisitation. It does
`this by paying particular attention to how previous pages are represented visually, and by integrating many
`revisitation capabilities into a single display space. Our preliminary evaluation of WebView indicates that users are
`enthusiastic about the functionality provided, and that it improves the efficiency of some navigational acts.
`1 Introduction
`Few, if any, technological advances have been as rapidly adopted as the World Wide Web. Fox (1999) notes that
`while radio and TV took thirty-eight and thirteen years respectively to attract 50 million users, the Internet took
`only four. User interfaces to web browsers have had little time for iterative refinement, and it should therefore be
`unsurprising that there are usability problems in even the most rudimentary facilities they provide.
`
`In our previous related work we have shown that page revisitation—the act of returning to previously seen pages—
`is a fundamental part of web navigation. About 58% of all pages a person visits are to ones they have seen before,
`and use of the browser’s Back button accounts for more that 30% of actions at the browser (Tauscher and
`Greenberg 1997). We have also shown that while many users have a naïve understanding of Back’s stack-based
`behaviour (Cockburn and Jones 1996), it is a reasonable (but not optimal) way to return to the most recently visited
`pages (Greenberg and Cockburn 1999). Returning to distant pages requires other revisitation schemes such as
`history lists and bookmarks. Many researchers believe these utilities could be improved and are now investigating
`alternate revisitation mechanisms.
`
`In this paper we describe our own revisitation system called WebView and its preliminary evaluation. WebView is
`an add-on window to Netscape Navigator that presents an automatically generated graphical overview of the user’s
`browsing paths. It provides a variety of facilities for navigational shortcuts, and it allows the user to tailor the
`display of a large set of pages. In designing WebView, we paid particular attention to two fundamental problems
`that all systems providing graphical navigational assistance must address. First, how can we assist the user in
`identifying previously visited pages, and second, what display organisation schemes can be used to enhance the
`visualisation of large sets of previously visited pages? The design issues associated with these two problems are the
`focus of a companion paper (Cockburn and Greenberg 1999).
`
`The structure of this paper is as follows. Section 2 describes the general design objectives that have motivated and
`guided our development of WebView. Section 3 describes WebView’s interface and its functionality. In Section 4
`we evaluate WebView, and find both quantitative and qualitative indications of its usability. Section 5 concludes
`the paper and provides directions for further work.
`2 Design Goals Motivating WebView
`2.1 High ratio of utility to screen real-estate
`
`Application windows that directly support a user’s primary task, tend to be continually present on the display
`during use e.g. word processors, web browsers, and spreadsheets. In contrast, secondary windows that augment the
`task are raised and hidden as required. The problem is that iconified windows that are running in the background
`require small levels of overhead to bring them to the foreground, and this may be sufficient to preclude their use if
`the value of the information they provide is low. The ‘History’ windows of current browsers such as Netscape and
`Microsoft Internet Explorer demonstrate this issue—despite being readily accessible through menu or icon
`selection, they are rarely used. Tauscher and Greenberg (1997) indicate that less than 1% of web-browser actions
`are on the history list, and Hightower et al. (1998) indicated that the value is less than 0.1%.
`
`APPLE 1024
`
`1
`
`

`

`Integrated navigation support for pages that are temporally near and far
`
`To overcome this problem, we believe revisitation systems must have high utility as well as compact
`representation. In our design of WebView, we do not expect users to have it continually on display. We do,
`however, intend that WebView should provide a sufficiently usable, efficient and visually compact system so that
`users are willing to un-iconify it for anything more than the shortest navigational path.
`2.2
`Current commercial browsers support a wide range of navigation support features, but they are spread across a
`wide variety of user interface components. Navigation features of Internet Explorer 5 and Netscape Navigator
`include Back and Forward buttons, pull-down menus off these buttons, ‘History’ windows, pull-down URLs
`off the ‘Address’ or ‘Location’ text-entry box, and ‘bookmarks’ or ‘favourites’. It is our intention that enhanced
`navigation systems such as WebView should integrate the functionality provided by these diverse features into a
`single unified system. A person should be able to use this integrated system to revisit pages regardless of whether
`they are temporally near (suitable for revisitation with the Back button) or far (requiring utilities similar to
`bookmarks or history).
`
`We believe that two system capabilities are required to achieve this goal of integrated navigation support. First, the
`system should automatically capture and record information about all pages that the user visits. This is needed
`because users often do not know that the page they are currently viewing will be important to them in the future.
`Analogous observations have been made in studies of users’ management of email (Mackay 1988), where the
`stereotypical behaviour of email ‘archivers’ is to keep
`copies of all email messages in case they become important
`at a later time. Furthermore, even when users do know that
`a page will be important to them later, they can forget to
`bookmark the page. By automatically capturing information
`about the page, the set of pages that the user must search
`through is enormously reduced (from the entire web to the
`set that the user has previously viewed).
`
`Table 1: Data on bookmark collections.
`Total
`Max.
`Num.
`Items / folder
`folders mean max.
`items
`depth
`113
`3
`18
`6
`21
`386
`4
`44
`8
`34
`541
`4
`85
`6
`27
`56
`5
`14
`4
`11
`104
`3
`13
`8
`18
`117
`3
`11
`10
`26
`85
`3
`4
`21
`80
`233
`3
`26
`8
`36
`204.4
`3.5
`26.9
`8.9
`31.6
` 26.4
`172.9
`0.8
`5.2
`21.2
`
`Mean
`S.D.
`
`Second, lightweight and implicit bookmarking techniques
`should be available to help the user return to important
`pages. The primary problems with current bookmarking
`schemes are that they require an explicit act from the user,
`and that they impose a management burden in reorganising
`their structure. Table 1 shows data that we collected on the
`bookmark structure of eight Computer Science academics (all of whom use Netscape Navigator). It is clear from
`the data that bookmarks require extensive management. The number of bookmarked items ranges from 56 to 541,
`with a mean of 204. All of the subjects had a hierarchical structure with at least three levels of depth. The right-
`hand column shows that the range for ‘Maximum items in a folder’ is from 11 to 80, with a mean of 31.6 (see also
`the bookmark study by Abrams et al. 1998). Bookmarks are normally accessed via pull-down menus, and despite
`research evidence that demonstrates that broad/shallow menu structures are preferable to narrow/deep structures
`(Landauer and Nachbar 1985), it is unlikely that menus with over thirty items in one level are efficient. Organising
`bookmarks is also problematic to the point that prior studies had described it as one of the top three Web usability
`problems, with users sometimes spending considerable effort re-organising them, or just deciding its not worth the
`effort (Abrams et al 1998). While bookmarks are valuable in principle, conventional browsers impose an excessive
`burden onto people wishing to use them: they require a user to decide upon a page’s importance ahead of time; they
`must be organised to be effective (especially when there are many of them); and they require housecleaning as old,
`unneeded bookmarks clutter the space.
`
`By automatically capturing information about every page the user visits, and by providing powerful searching and
`filtering schemes, we believe that systems can largely negate the need for explicit bookmarks. Furthermore, access
`to the user’s most important pages can be enhanced by heuristics that infer information about the page through data
`such as visit counts. Finally, lightweight interface mechanisms that allow the user to explicitly mark a page as
`important should be provided.
`2.3 Effective support for page identification, display organisation, and page retrieval
`The previous two design goals raise conflicting requirements for systems that integrate techniques for web
`navigation. They must maximise the richness of the information display, and yet they must allow the user to view
`and access all previously visited pages, some of which will be unimportant to the user. To attain an acceptable
`compromise between information density and information overload it is necessary for systems to support the user
`in three ways.
`
`Support for page identification. Any graphical scheme for page revisitation must provide a visual representation
`that aids page identification. The most common technique is to use the <Title> text extracted from the page’s
`HTML. In our companion paper (Cockburn and Greenberg 1999) we described the severe limitations of
`
`2
`
`

`

`representing pages with only the <Title> text, the consequence being that people can have problems finding a
`particular page in a list. The companion paper also described several alternative schemes for representing pages;
`the one used in WebView (Section 3) is to provide several redundant cues to page identification, including
`zoomable thumbnail images, a ‘dogears’ metaphor for the page visit count and for bookmarks, the page’s <Title>
`text, and the page URL.
`
`Effective display organisation. It is essential that the visualisation of a large set of pages is organised in a manner
`that aids browsing, searching and categorisation. Yet there is no easy answer on how best to do this: major issues
`are discussed in the companion paper (Cockburn and Greenberg 1999), and a survey of display organisation
`schemes is given in Cockburn and Jones (1997). WebView, described in Section 3, supports two of the more
`promising display organisation schemes: a temporal view, and a ‘hub and spoke’ view.
`
`Implicit and explicit bookmarking through dogears
`
`Powerful search facilities. Effective mechanisms for page identification and display organisation can increase the
`number of pages that the user can interact with and perceive in the display, but searching and information filtering
`schemes will be necessary to help the user restrict the number of pages displayed. As well as text-based searches
`for title and URL information, we also wish to investigate the usability of searches that are based on abstract page
`properties such as the page visit count, and the timing of the first and last visit to the page.
`3 The WebView Prototype
`WebView is an add-on window that interacts with unaltered versions of Netscape Navigator1. Whenever the user
`visits a page in Netscape, WebView’s display is automatically updated to reflect the action. As with conventional
`systems, clicking on the text-title alongside any page makes Netscape navigate to the page. The following
`subsections describe WebView’s techniques for page identification and page organisation.
`3.1 Representing pages
`WebView captures and displays a miniaturised zoomable thumbnail image of the rendered page (Figures 1-2), an
`approach also used in a few other research systems (Hightower et al. 1998; Ayers and Stasko 1995). It also detects
`the title and URL of the page, and these are (optionally) displayed alongside the thumbnail. Because some
`thumbnails may be difficult to distinguish from others (such as a site’s pages that follow a standard look, we also
`provide larger views: mousing over any miniaturised thumbnail causes it to zoom to approximately four times the
`size (Figure 2a bottom).
`3.2
`WebView combines thumbnails with bookmarking cues through a ‘dogears’ metaphor (Figure
`1). Dogears encode information about the number of visits to a page (an implicit bookmark
`based on the idea that frequently visited pages are somehow more important than others) and
`about page bookmarks (where a person marks a page explicitly). An implicit bookmark
`displayed as a ‘dogear’ at the top-left of each page grows progressively denser green with each
`visit to the page, allowing users to readily identify pages that they visit frequently. An explicit
`‘bookmarking’ dogear is added to the bottom left hard corner of the thumbnail when the user
`clicks it with the middle mouse button.
`3.3 Organisation schemes: ‘Hub-and-spoke’, temporal display, and shortcut menus
`WebView supports two primary display organisation schemes, controlled by options under its ‘View’ menu. Its
`‘hub-and-spoke’ view (Figure 2a), displays the tree-like nesting relationship between the storage location of pages.
`In Figure 2a, for example, it is clear that the user has visited four pages under the “Academic Staff” page.
`Organising the display as a structural tree, such as this, is intended to help the user see the context of their
`navigational acts: pages on a similar topic are likely to be displayed in close proximity to each other. To avoid
`excessive nesting depth in the display, each new web site that the user visits is added to the display at the top-level
`of the nesting structure. The figure therefore shows that the “COSC Home Page”, “University of Canterbury…”
`and “Thimbleby’s home page” are each located at separate web-sites. Naturally, separating sites in the display will
`also separate pages on related topics that are stored at different sites. To ease this problem, WebView adds a
`feature that displays cross-site navigational links. The arrowhead links connecting thumbnails in Figure 3 shows
`that the user navigated from the “Academic Visitors” page to “Thimbleby’s home page”, and then to the page
`“Professor Ian. H. Witten…”, each of which are stored at different sites.
`
`Figure 1: Dog-eared
`thumbnail.
`
`
`1 Interaction with unaltered versions of Netscape is achieved through Netscape’s remote control facilities
`(home.netscape.com/newsref/std/x-remote-proto.html).
`
`3
`
`

`

`WebView’s ‘temporal’ display organisation,
`shown in Figure 2b, provides a complete
`temporally ordered recency list of pages with
`duplicates removed. The back and forward
`buttons at the top of WebView’s display
`operate on
`this
`temporally ordered data
`structure, and the ‘Up’ button has the same
`effect as the stack-based navigation scheme
`currently used by Netscape Navigator and
`Microsoft Internet Explorer (which provides
`an incomplete recency list). Greenberg and
`Cockburn (1999) provide details of the design
`issues and algorithm used in WebView’s
`temporal display.
`
`Finally, WebView parses the page’s HTML
`contents, extracts the page links, and uses this
`information to construct a pop-up ‘shortcut’
`menu associated with the thumbnail (Figure 4).
`Clicking on a thumbnail pops up the shortcut
`menu that can be used to navigate directly to
`any subordinate child page. The shortcut menu
`works within both temporal and hub-and-
`spoke views.
`4 Preliminary Evaluation
`We performed a preliminary evaluation of
`WebView to gain insights into its usability and
`efficiency,
`and
`to
`direct
`subsequent
`development work. We used seven volunteer
`subjects, all graduate Computer Science
`students or tutors. All of the subjects used
`web-browsers as part of their daily work (6
`using Netscape Navigator and one using
`Microsoft Internet Explorer), and all were very
`familiar with the structure and contents of the
`web site used for the evaluation tasks (shown
`in Figure 5). Given the preliminary nature of
`the evaluation, we wanted to see if it worked
`with knowledgeable ‘resilient’ users, and thus
`accepted obvious bias in our subject group.
`
`a) hub and spoke
`
`b) temporal ordering
`
`Figure 2: WebView's two display organisation schemes.
`
`Figure 3: Cross-site link
`indicators
`
`Figure 4: Shortcut pop-up
`menu
`
`Each subject participated in a twenty-minute evaluation that compared
`WebView’s navigation efficiency (in terms of task time) with that of Netscape.
`Only WebView’s temporal view was used during the tasks (Figure 2b); we will
`evaluate the ‘hub and spoke’ view in subsequent studies. Subjects each received
`a five-minute introduction to the system. They then carried out two navigational
`tasks, each with two parts. A stopwatch was used to time all tasks. After the
`tasks, they answered three questions and provided general comments.
`
`Task 1 compared Netscape’s support for navigation with that provided by
`WebView. In Task 1a, without a WebView client running, the subjects were
`instructed to navigate to the “Teaching” page. At this point they were instructed
`to “Visit Andy’s page, and then Wal’s page, as quickly as possible”. They were
`asked to confirm that they understood the task, and then they were told to start. The stopwatch was stopped when
`they reached “Wal’s” page (assuming they had previously displayed “Andy’s” page).
`
`Figure 5: Web pages used
`in the evaluation tasks.
`
`In Task 1b, a new Netscape session was started, and a new WebView client was launched i.e., only the root page
`was included in WebView’s list. The subjects were asked to complete an identical navigation task to 1a (to visit
`“Andy’s” then “Wal’s” page from the “Teaching” page), but they were instructed to do so using only the
`navigational features provided by WebView. Although WebView is primarily designed to support page
`revisitation, its shortcut menus and other capabilities allow a wide range of navigational actions. We felt that this
`task would provide insights into the usability of many of WebView’s capabilities, even though the task is outside
`
`4
`
`

`

`its primary usage scenario. The risk of a learning effect between tasks 1a and 1b was minimised by having the
`subjects navigate through the test web pages during the pre-test introduction.
`
`Table 2: Evaluation results: timings for Tasks 1 and 2 (seconds)
`
`Subject
`
`1
`2
`3
`4
`5
`6
`7
`Mean
`S.D.
`T Test
`
`Task 1b
`Task 1a
`(WebView)
`(Netscape)
`30.2
`14.2
`26.6
`22.4
`18.0
`15.4
`31.9
`20.0
`21.0
`68.0
`22.4
`15.4
`58.3
`20.6
`29.8
`25.1
`13.5
`19.2
`t(6)=0.48, p = 0.65
`
`Task 2b
`Task 2a
`(Netscape)
`(WebView)
`6.8
`3.8
`7.3
`3.0
`4.5
`1.0
`28.9
`6.5
`7.0
`1.0
`8.4
`1.3
`4.6
`1.4
`9.6
`2.6
`2.0
`8.6
`t(6)=2.71, p = 0.035
`
`Table 3: Questionnaire and responses
`
`Responses using a five point Lickert scale: from 1 “not at all” to 5 “very”
`
`Subjects:
`
`How useful were thumbnail images
`in recognising pages?
`
`1
`
`4
`
`2
`
`2
`
`3
`
`2
`
`4
`
`2
`
`5
`
`3
`
`6
`
`2
`
`7
`
`4
`
`In Task 2a, the subjects were asked to use
`WebView to return to the “Teaching” page
`from “Wal’s” page, and in Task 2b they
`performed the same action using Netscape.
`This revisitation task is close to WebView’s
`primary usage scenario.
`
`Subjects then answered the questions in
`Table 3.
`4.1 Observations and Results
`Five of the seven subjects were enthusiastic
`about the system, three of them extremely so,
`making statements such as “It’d be great to
`have a system like this.” The two subjects
`who were not enthusiastic about the system
`were primarily
`concerned
`about
`the
`redundancy
`introduced by having
`two
`different ways of navigating (the “Netscape
`way” and the “WebView way”). This is not a
`problem, as the ultimate
`intention with
`systems such as WebView is that they should
`replace, rather than duplicate, the current set
`of navigational facilities.
`
`How usable were the link shortcuts
`menus off the thumbnails?
`
`If available, how likely would you
`be to use a system like WebView?
`
`2
`
`2
`
`4
`
`4
`
`3
`
`4
`
`5
`
`5
`
`5
`
`5
`
`5
`
`5
`
`3
`
`2
`
`Users readily adapted to most of WebView’s
`navigational facilities. For instance, in Task
`1b, when navigating to “Wal’s” page from
`“Andy’s” page, all users
`immediately
`selected
`the
`shortcut menu off
`the
`“Academic Staff” thumbnail without returning to the “Academic Staff” page. Table 2 shows that there were no
`significant differences in the mean times for task completion in Task 1a (Netscape) and Task 1b (WebView) (two-
`tailed paired T-Test, p > .1). This lack of difference is good news, for although WebView was designed to favour
`page revisitation, this indicates that people can use it to navigate to new pages as well as old ones without
`excessive penalty.
`
`In Task 2a all users were significantly faster in using WebView to return to the “Teaching” page from “Wal’s”
`page. The mean times for Netscape and WebView were 9.6 seconds and 2.6 seconds respectively, giving a
`statistically significant improvement (t(6) = 2.7, p < .05, two-tailed paired T-Test).
`
`Table 3 shows that subjects 2 to 6 were enthusiastic about the system, responding positively to questions two and
`three. Interestingly, the only subjects who thought the thumbnail images were useful in recognising pages were the
`ones that rated the system poorly (1 and 7). When asked for comments about the thumbnails, several subjects stated
`that the thumbnails would probably be more useful when navigating through sites that they were unfamiliar with.
`Subject 4 stated that he had not found the thumbnails useful because the tasks were introduced verbally, using the
`same names for the pages as was displayed in the text associated with each page.
`
`We observed two primary usability problems with WebView. First, in Task 1b when using WebView’s pop-up
`shortcut menus, the subjects often took time searching through the menu for the desired link. In contrast, when
`using Netscape in Task 1a, the users knew immediately where the desired link would be displayed. There are no
`obvious ways to ease this problem, but we are not overly concerned about it. The pop-up menus provide powerful
`shortcut facilities that promise to increase the efficiency of navigation, and we view the time taken to search for
`links in the menu as a necessary design compromise in providing these facilities. We also suspect that the subjects
`would have spent more time searching for links in Netscape if they had not been so familiar with the web space
`used in the study.
`
`A second problem affected subject 7 severely (also noted by subject 6 during the pre-test introduction). In Task 1b,
`subject 7 repeatedly accessed the wrong shortcut menu: for example, he popped up the “Homepage” menu several
`times when he needed to access the “Academic Staff” menu. In comments after the task he reported that he found
`the movement of thumbnails extremely confusing, and that he wanted the thumbnails to “stay in the same place”
`(WebView automatically moves the current page to the top of the display to maintain its temporal list of pages). He
`
`5
`
`

`

`stated that the ‘hub and spoke’ view (demonstrated during the introduction) would have been much easier to use
`“because it shows where you are”.
`5 Conclusions and Further Work
`Our work with WebView is investigating interfaces that provide access to a wide range of techniques for revisiting
`web pages. The ultimate aim of the research is to develop interfaces that improve the usability and efficiency of
`navigation with current commercial browsers, while simultaneously rationalising the diverse and seemingly ad-hoc
`range of features that they offer. Although in the early stages of development, WebView provides powerful
`facilities for page identification and display organisation. These include zoomable thumbnail images of visited
`pages, shortcut link menus, a dogears metaphor for bookmarking and page identification, and a tailorable scheme
`for displaying pages either by temporal properties or by ‘hub-and-spoke’ structural relationships.
`
`A preliminary evaluation indicates that WebView can provide statistically significant improvements in efficiency
`in some navigational tasks. The usability subjects were also generally enthusiastic about the support it provided.
`While there are clearly things we did not evaluate—implicit bookmarks, dogears, the hub and spoke view, complex
`revisitation sequences—the positive results encourage us to continue our development and study of WebView.
`
`Our further work will focus on two areas. First, we will extend WebView’s functional capabilities. This work will
`focus on extending the temporal range of pages displayed, and on dynamic queries for searching and filtering the
`page display (Greenberg and Cockburn, 1999). In extending WebView’s temporal range, we will experiment with
`expandable temporal segments such as ‘last week’ in a similar manner to that currently used in Microsoft Internet
`Explorer’s ‘History’ feature. The second main area for further work will be on evaluation, with a focus on detailed
`studies of all its features, as well as longitudinal studies of WebView’s use over time and over various browsing
`tasks.
`
`Acknowledgements. Partial funding for this work was provided by NSERC Canada and Microsoft Research.
`References
`1. Abrams, D., Baecker, R. and Chignell, M. 1998. Information archiving with Bookmarks: Personal Web Space construction
`and organization. In ‘Proceedings of CHI’98’, 41-48.
`2. Ayers, E. and Stasko, J. 1995, Using graphic history in browsing the World Wide Web, in ‘Proceedings of the Fourth
`International WWW Conference. 11–14 December, Boston’.
`3. Cockburn, A and Greenberg, S. 1999, Beyond the Back Button: Issues of Page Representation and Organisation in Web
`Navigation Tools, Companion paper. Submitted to OzCHI’99.
`4. Cockburn, A. and Jones, S. 1996, ‘Which way now? Analysing and easing inadequacies in WWW navigation’,
`International Journal of Human-Computer Studies 45(1), 105–129.
`5. Cockburn, A. and Jones, S. 1997, Design issues for WWW visualisation tools, in ‘Proceedings of RIAO’97, Montreal’, 55–
`74.
`6. Fox, R. 1999, ‘NewsTrack’, Communications of the ACM 42(5), 9.
`7. Greenberg, S. and Cockburn, A. 1999, Getting back to back: Alternate behaviors for a web browser’s back button, n
`‘Proceedings 5th Conference on Human Factors and the Web’. www.nist.gov/hfweb/
`8. Hightower, R., Ring, L., Helfman, J., Bederson, B. and Hollan, J. 1998, Graphical multiscale web histories: A study of
`padprints, in ‘Proceedings of Hypertext’98’, ACM Press, 58–65.
`9. Landauer, T. and Nachbar, D. 1985, Selection from alphabetic and numeric menu trees using a touch screen: Breath, depth,
`and width, in ‘Proceedings of CHI’85’, ACM Press, 73–78.
`10. Mackay, W. 1988, More than just a communication medium: Diversity in the use of electronic mail, in ‘Proceedings of
`CSCW’88, Portland’, 344–353.
`11. Tauscher, L. and Greenberg, S. 1997, ‘How people revisit web pages: Empirical findings and implications for the design of
`history systems’, International Journal of Human Computer Studies 47(1), 97–138.
`
`6
`
`