WebComets: A Tab-Oriented Approach

for Browser History Visualization

Daniel Cernea

1,2

, Igor Truderung

, Andreas Kerren

and Achim Ebert

Computer Graphics and HCI Group, University of Kaiserslautern, P.O. Box 3049, D-67653 Kaiserslautern, Germany

Computer Science Department, ISOVIS Group, Linnaeus University, Vejdes Plats 7, SE-35195 Växjö, Sweden

Keywords: Browser History Visualization, Glyph-based Techniques, Visual Knowledge Discovery, Time Series,

Parallel Browsing.

Abstract: Web browsers are our main gateways to the Internet. With their help we read articles, we learn, we listen to

music, we share our thoughts and feelings, we write e-mails, or we chat. Current Web browser histories

have mostly no visualization capabilities as well as limited options to filter patterns and information.

Furthermore, such histories disregard the existence of parallel navigation in multiple browser windows and

tabs. But a good understanding of parallel browsing behavior is of critical importance for the casual user

and the behavioural analyst, while at the same time having implications in the design of search engines,

Web sites and Web browsers. In this paper we present WebComets, an interactive visualization for extended

browser histories. Our visualization employs browser histories that capture—among others—the tab-

oriented, parallel nature of Web page navigation. Results presented in this paper suggest that WebComets

better supports the analysis and comparison of parallel browsing and corresponding behavior patterns than

common browser histories.

1 INTRODUCTION

Since the release of the World Wide Web to the

public, the Internet instantly became an important

source of information as well as a communication

platform without which today’s world is hard to

imagine. The number of registered domains

increased in the last decade from 35 to

approximately 350 million, and the trend is

accelerating rapidly.

While different approaches have been developed

to navigate this mass of information, almost all of

them are still based on the well-known Web

browser. Through the browser, users search for a

particular piece of text or information, they work on

or through the Internet, they stream entertainment

media or communicate (Kleek et al., 2010). As a

result, the logging features of our browsers—called

Web browser histories—have the ability to reflect

our requirements, interests and activity. Therefore, it

seems that an efficient examination of our browser

histories offers the opportunity of retrospectively

inspecting users’ behavior on the Web.

But in recent years, the paradigm of Web

navigation has shifted and many browsers started

implementing “tabs” to support parallel browsing.

Tabs allow users to access and explore multiple Web

pages simultaneously. While similar, this experience

is still significantly different from the one of using

multiple browser windows, as the more lightweight

tabs allow users to have an overview of the opened

Web pages and even organize (e.g., group) the

various tabs by the topic of the loaded page. The

importance of tab-based operations can be further

supported by the work of Miyata et al. (Miyata and

Norman, 1986), where the presence of foreground

and background tasks and their interconnection in

the human mind is emphasized from the perspective

of cognitive psychology. As such, Web browser tabs

are specifically designed to follow this principle and

allow users to distribute their attention based on this

model.

As a matter of fact, a recent paper (Huang and

White, 2010) highlights that 57.4% of Internet

sessions in the browser make use of tab-based

parallel navigation. Other findings comprise that

these values fluctuate between 4-85% (Viermetz et

al., 2006). While not conclusive, such a wide range

suggests an even more important aspect: there is

currently only limited information and insight into

439

Cernea D., Truderung I., Kerren A. and Ebert A..

WebComets: A Tab-Oriented Approach for Browser History Visualization.

DOI: 10.5220/0004226004390450

In Proceedings of the International Conference on Computer Graphics Theory and Applications and International Conference on Information

Visualization Theory and Applications (IVAPP-2013), pages 439-450

ISBN: 978-989-8565-46-4

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

the way users explore and organize their parallel

browsing experiences, and a visualization tool would

be needed that is capable of capturing and reflecting

the intricacies of online navigation today. Such a

powerful representation enabled by meta-

information about user online sessions, has the

potential to simplify the tasks of researchers in fields

like information retrieval and behavioral sciences, as

well as offer an alternative to the casual user who

wants to explore his online multitasking and

browsing habits.

Sadly, in most cases the representation of the

history data is limited to a textual list of Web site

names and URLs that is sorted based on a certain

criterion, e.g., chronological or by load frequency.

Additionally, these browser histories give little

insight in the call hierarchy of the Web pages or the

relevance of a particular site to the users. In other

words, while browsers offer support for parallel

browsing, most current histories fail to capture this

aspect (Huang and White, 2010) and thus do not

reveal any interconnection patterns between the Web

pages or user sessions. Thus, as temporal features

are not sufficiently highlighted in conventional

histories, one cannot recognize the connections

between Web sites and browser windows and tabs.

This information is relevant in a setting where users

now engage in parallel navigation and distributed

attention between many open browser tabs and

windows.

In this paper, we address the problem of

designing an interactive visualization tool for

Internet browser histories that supports intuitive

search operations based on content and context

information, and that allows the tool user to more

quickly find, compare and analyze parallel

navigation behavior based on a set of existing—e.g.,

those described in (Huang and White, 2010)—and

novel metrics. The remainder of the paper is

organized as follows. First, we focus on research

that is relevant to the topic at hand and continue with

a requirement analysis. This is followed by a

detailed discussion of the data gathering process

together with the design decisions and the featured

interactions of our proposed visualization. In order

to validate our approach, we then describe an

evaluation of our tool. Finally, we conclude and

provide future research directions.

2 RELATED RESEARCH

Maybe the most common approach for visually

encoding browsing histories are tree representations.

Tools like MosaicG (Eric et al., 1995), PadPrints

(Hightower et al., 1998), Organic Bookmark

Management (Shen et al., 2011), WebMap (Doemel,

1994) and Domain Tree Browser (Gandhi et al.,

2000) use one or multiple vertical or horizontal 2D

trees to represent the domain-structure of the

navigated Web sites. In some cases, these tree views

are coupled with additional list views that highlight

the temporal order of visitation, as the tree

representations do not reflect the temporal

succession of events. Additionally, in many cases

screenshots of the Web pages are used as thumbnails

embedded in the nodes to support the recognition

process (Eric et al., 1995); (Hightower et al., 1998);

(Shen et al., 2011). Still, all these approaches

represent a Web page only once in the tree, even if it

is visited multiple times.

An alternative 2D graph representation focuses

on capturing and visualizing the branching events in

the navigation path (Waterson et al., 2002). These

visualizations manage to capture the sequential

aspect of the browsing process, as each accessed

page is drawn as an additional node in the graph. If

the user navigates back and accesses a different Web

site, the resulting branch will be accordingly

represented in the visualization.

A slightly different 2D space-filling solution is

offered by the Trails plug-in (Yu and Ingalls 2011)

that supports a hierarchical, chronological and

group-based representation of the visited pages.

Furthermore, it offers a statistical overview of the

most often visited Web sites. Another method for

representing browser histories is highlighted by

solutions that employ one (Kaasten and Greenberg,

2000) or multiple (Cockburn et al., 2003)

interconnected linear views that are enhanced by

graphical elements (e.g., thumbnails).

Besides 1D and 2D solutions, Web browser

histories have been developed that employ multiple

dimensions or intuitive metaphors. VISVIP (Cugini

and Scholtz, 1999) is a 3D representation of a

navigation log, where two dimensions are used for

drawing the Web site structure, while the additional

third one encodes the temporal information. On the

other hand, the combo WebBook and WebForager

(Card and Robertson, 1996) use the concept of a

book to give an overview of the Web sites as well as

offer an intuitive information-space for the user.

A special class of browser histories is

represented by the statistical summary histories.

Tools like SlifeWeb (Slife Labs, 2012), RescueTime

(Rescuetime, 2012) or Eyebrowse (Kleek et al.,

2010) are mainly focused on time management and

analytics, and allow users to generate their own

IVAPP2013-InternationalConferenceonInformationVisualizationTheoryandApplications

440

statistic view about how they—or others—navigate

the Internet.

However, browser histories are not the only type

of data revolving around complex, interconnected

temporal events. Other time-series visualizations

employing similar visual concepts to our approach

include World Lines (Waser et al., 2010), a

visualization technique for exploring the alternative

paths of heterogeneous simulation runs, and

LeadLine (Dou et al., 2012), a visual analytics tool

for identifying and representing meaningful events

in news and social media data.

While diverse and functional, none of these

methods focuses on the complex parallel browsing

habits of today, where tabs and windows have

become means for the user of organizing his

thoughts, actions and accessed information (Huang

and White, 2010). The importance of a tool for

visualizing, analyzing and comparing parallel

browser behavior is further highlighted in (Aula et

al., 2005), because users tend to use multiple

windows and tabs as means for backtracking (e.g.,

users abandon the use of in-browser back operations

in favor of opening new tabs and switching between

them) and multitasking (e.g., users interact with one

tab while Web pages are being loaded and processed

in others). Similarly, findings from (Spink et al.,

2006); (Weinreich et al., 2006) suggest that users

often employ parallel browsing in Web search tasks

for reasons like comparing search results, executing

multiple queries, interacting with a page while

others are being loaded, etc.

3 REQUIREMENT ANALYSIS

To support flexible search and analysis efforts, all

control and interaction elements that users employ to

organize complex and parallel browsing sessions—

such as windows–tabs and back–forward operations

—need to be recorded and graphically represented in

a first step (cp. Subsection 4.1). The captured data

organized into a set of different user profiles will be

at the core of the later visualization. It has to embed

rich meta-information that could be of interest for

the user of WebComets, i.e., for researchers in

information retrieval, behavioral sciences and

related fields (called tool user or analyst in this

paper). Following the study described in (Huang and

White, 2010), typical research questions for a better

understanding of parallel browsing behavior on the

Web are for example: When and to what extent are

users parallel browsing on the Web? Or what affects

parallel browsing behavior during interaction with

Web search results?

Another possible research question is to identify

reasons why users revisit pages—for example,

because of monitoring pages (Kellar et al., 2007);

(Adar et al., 2008) — and how this is typically done

in a multi-tab browser environment. For supporting

such studies and for finding answers for such

questions, a visualization tool has to offer specific

functionalities. We have composed a list of

requirements that need to be satisfied by a browser

history visualization based on preliminary studies,

user feedback about browser histories and their

limitations, and information about the nature of

parallel browsing behavior from the previously

referenced publications. Fundamental requirements

for the visualization of the captured data are:

• The visualization should offer an overview of the

loaded data and support detailed investigations. This

can be achieved by means of tailored interactions

and methods like “detail-on-demand”.

• The temporal flow of the visualized navigation

sessions needs to be clearly distinguishable. While

most Web browsers only display a chronologically

sorted list of the accessed Web pages, it is important

that the temporal dependencies are visually and

relationally highlighted.

• For each visited Web page, additional meta-data

needs to be captured like the duration of each visit or

the duration of interacting with the Web page. While

many solutions already count the number of

executed accesses to each Web page, it is important

to visualize the temporal sequence of events related

to any sequence of navigated pages because

documents may be inspected more often and for

longer periods.

• Web site categories, like search engines or news

Web sites, should be introduced and represented.

Current browsers support bookmarking of Web

pages, which implies saving the Web page’s address

while at the same time tagging it with the help of

keywords or categories. These operations increase

the retrievability of stored bookmarks and should

also be offered by our system.

• The visualization should clearly represent which

navigation path—or sequence of visited Web

pages—the users have followed during their

browsing session. This includes information about

what browser windows and tabs have been opened

and closed, and how these are connected to

particular accessed documents. Thus, it will be

possible to reconstruct the steps that lead to the

opening of a particular Web page—a missing feature

in many related works.

WebComets:ATab-OrientedApproachforBrowserHistoryVisualization

441

Requirements that improve scalability (with

respect to log size and number of user profiles) and

analysis possibilities:

• Visualizing multiple browsing histories at the same

time should be supported in order to allow

comparison and analysis operations (e.g., detect if

multiple users have similar interests or if users have

communicated in a certain time period). This feature

is mostly not supported so far.

• Connections between similar Web sites should be

emphasized as these might be relevant alternatives in

search and analysis tasks.

• Our tool users should be able to search for

particular Web pages based on content (e.g., title or

category) and context information (e.g., approximate

duration of access or pages accessed prior to the one

in question). While content-based search is present

in all history lists, a search for the context is not

supported in most cases, as this requires additional

information about the navigation paths the user

followed.

• Equally important for the analysis of parallel

browsing behavior is the search for navigation

patterns, i.e., finding specific structures (motifs) in

the navigation graph which results from branching

out from a linear navigation behavior by using tabs

or additional browser windows.

4 WebComets

WebComets is a system for the interactive

visualization of extended, tab-based browser

histories. It was implemented in Adobe Flash

ActionScript, with both online and standalone

capabilities. The representation and interaction

metaphors it incorporates satisfy the requirements

highlighted in the previous section. Figure 1 shows a

screenshot detail of our tool. To achieve the required

functionality and analysis capabilities, WebComets

cannot solely rely on information gathered by

standard logging systems. Thus, we propose an

extended logging mechanism as described in the

next subsection.

4.1 Extended Browser History

Some of the previously highlighted requirements

cannot be achieved with the history features of

major Web browsers. For example, browsers like

Mozilla Firefox or Internet Explorer do not record

the duration for which a user has actively interacted

with a Web page. More importantly, browsers do not

focus on capturing the parent-child relationships

between accessed Web pages and even less the

connections between opened tabs or windows. Other

Figure 1: WebComets visualization of the parallel browsing histories of two users (light-brown and light-grey background).

Each horizontal line represents the timeline of a tab that the user has opened, while vertical branches highlight new tabs that

have been created by clicking a hyperlink in the parent tab. The comet glyphs encode loaded Web sites and their color

coding represents topics. Their position on the time axis depends on the moment when they were accessed. In the current

visualization, one can notice that the first user had up to six tabs open within the shown time. The second user employed

two browser windows without additional tabs and spent most of his/her time on one Web page as represented by the long

comet tail in the second window.

IVAPP2013-InternationalConferenceonInformationVisualizationTheoryandApplications

442

researchers have also encountered this difficulty

when investigating the parallel browsing behavior of

Internet users (Huang and White, 2010). The

unavailable information included, among others,

missing source tabs for branching operations and no

information on how a tab or window was created

(new tab, hyperlink click).

To address this, we developed a Mozilla Firefox

browser add-on with the help of Javascript and

libraries like jQuery and Kinetic. The add-on

incorporates the ability of recording and saving an

Extended Browser History (EBH) inside an SQLite

database. The information collected like this can be

subsequently accessed and visualized for multiple

browsing sessions and users.

For any current user profile, the extension saves

the navigated URLs together with relevant

additional information. The complete EBH includes

a subset of the metrics employed in (Huang and

White, 2010) and a set of additional metrics relevant

to the analysis and comparison of parallel browsing

habits and behavior. Thus, for each user profile, the

EBH records the following information:

• user profile information (such as username),

• opening and closing times for each tab and window

(tab and window sessions), as well as

• type of creation for each tab and window, i.e.,

opened blank or through a link from another tab /

window (branching).

To complement this, the following data will be

recorded for every accessed Web page:

• The title of the Web page including its URL.

• Information about how a Web page was opened:

through direct input of the URL in the address bar

by the user, through clicking of a link in another

Web page, or through the browser’s integrated

back/forward operations.

• Category of the Web page based on a static list of

Web domains.

• Number of accesses to a domain and a particular

URL (pageview).

• Additional time intervals: total time – the time

interval when a Web page was loaded and discarded;

focus time—the time interval for which a Web page

was in the foreground; active time—the time interval

for which the user interacted with a Web page.

Note that the tab switches metric proposed by

Huang et al. is currently not being stored in the

EBH, as we argue that additional time intervals offer

an alternative view for the distribution of the user’s

attention over multiple tabs and windows. In the

following, we highlight the visual design and

interaction capabilities of WebComets together with

supplementary information about the EBH.

4.2 Design

In order to satisfy the requirements highlighted in

Section 3, the WebComets visualization has to

consider a variety of aspects. Probably most

important is the representation of the temporal

dimension and the mapping of the Web pages to a

time axis. In order to use the larger width to height

ratio of modern screens (widescreen), a visualization

concept was devised that maps the timeline to the

horizontal axis, from left to right. In this

representation, each accessed Web page is displayed

as a circle and gets assigned its corresponding

position on the timeline (x-axis).

The encoding of the parallel navigation that the

users are involved in by using multiple browser

windows and tabs is another important aspect of the

visualization. WebComets represents each browser

tab as a separate horizontal line segment that is

parallel to the time axis (Figure 2). This combination

of patches and parallel segments is similar to the

representation of a parallel browsing session in

(Huang and White, 2010), as well as to (Krstajic et

al., 2011) where multiple time-series are visualized

through a comparable solution.

Figure 2: Rectangular representation of two browser

windows and their corresponding tabs. The different

background colors for the rectangles suggests that the

browser windows were created by two user profiles.

As tabs can be opened manually or by clicking a

link in another tab, this can result in a tree-like

structure that also suggests connections in terms of

hyperlinks, but possibly also themes between

various Web sites. This parent-child relationship is

represented in the visualization as two horizontal

lines connected by a vertical one. At the same time,

if the user opens a tab manually, there is no clear

way of connecting the first Web page of this tab to

any other already open pages. Therefore, a new tab

line is shown as disconnected from the rest of the

WebComets:ATab-OrientedApproachforBrowserHistoryVisualization

443

tabs that were already loaded (Figure 2).

At the same time, multiple opened browser

windows are visually encoded as framed rectangular

areas, where each rectangle contains a tree-like

structure of tabs that reflects the opened tabs in each

window during the user session, see Figure 2. As the

rectangles stretch along the horizontal axis, its left

and right margins represent the opening and closing

times of the window. Note that a rectangular

representation can be also activated at the tab-level

to enforce the navigation patterns of the users. All

rectangular shapes have a specific background color

that identifies them as windows/tabs belonging to a

certain user profile.

4.2.1 Visual Encoding

The representation of the visited Web pages has at

its core a glyph enriched with graphical elements

encoding multiple EBH attributes, as shown in

Figure 3. These circular glyphs are mapped on the

horizontal axis to the moment in time when the

corresponding Web page was loaded, while the

vertical positioning identifies the tab in that the Web

page was accessed.

Figure 3: Circular glyph representation of a visited Web

page. The figure highlights two versions of representing

the focus and active times of the visited Web pages: as a

comet tail (top) or as beams (bottom). Active time is

represented in light-blue, whereas focus time is in dark-

blue.

In a browser history, there are two major elements

that suggest the importance of a Web page to the

user: the number of visits (pageviews) and the time

the user has spent actually interacting with the page

or leaving it active. As such, we encoded these two

metrics with visual attributes that would reflect their

importance through size and shape, according to the

separability of dimensions (Ware, 2004). The radius

of each circular representation encodes the number

of visits the user executed in the current session to a

particular domain, e.g., “www.google.com”. As a

result, pages that are part of domains visited more

often will have a larger circular glyph than those that

have domains visited rarely.

Tool users have the option to activate a pie chart

representation for all glyphs, that captures the ratio

of visit counts for the current Web page compared to

the overall visit count for the domain. For example,

if the domain “google.com” has been accessed six

times in total and the current Web page

(“www.google.com/search?q=conference”) only two

times, then the pie chart will encode a sector of 1/3.

This type of information can supply vital insights to

the analysts, as in many browsing instances multiple

pages of a certain domain are accessed, but only

some are accessed more often or for a longer period

of time.

As mentioned in Subsection 4.1, EBH includes

three time intervals for every Web page: total, focus

and active time. Remember that focus time stores the

amount of time the Web page was in the foreground

and active time captures the duration for which the

user was clearly present and interacting with the

Web page, e.g., by mouse movements or key

strokes. Therefore, the following relationship is

valid for any visited page:

total_time ≥ focus_time ≥ active_time

(1)

As the horizontal axis is mapped to temporal

features, it is thus intuitive to encode the total time

for a Web page as being represented by the

horizontal segment between its glyph representation

and the position of the glyph for the following Web

page on the same tab. Additionally, the focus and

active times are visualized as subsegments of this

total time. This is achieved by two alternative

representations (cf. Figure 3): the first one has the

shape of a comet tail, and the resulting glyph

metaphor is meant to suggest the direction of the

temporal flow; the second one is a beam-based

representation that is more compact, and thus used

on collapsed branches and space-saving

representations (see Section 4.3). Both intervals

have their origin at the loading time of the current

Web page, and their length is proportional to the

total time the user spent on the page.

IVAPP2013-InternationalConferenceonInformationVisualizationTheoryandApplications

444

4.2.2 Web Site Categories and Browsing

Issues

For both encodings, the glyphs are colored based on

a correspondence to Web site categories. The

categories support the quick differentiation and

detection of various activities that the browser users

were involved in. They have been generated based

on (Gross, 2004); (Katz and Aspden 1997), where

common in-browser activities are highlighted. For

each visited link, a category is selected by the

Firefox extension based on the domain of the current

Web site. The process is static and employs a

database of domains that is divided into 15

categories. The lists of relevant domain names for

each category have been generated with the help of

the Alexa Top 500 Web sites database (Alexa,

2012). If a domain is not found in the database, the

Web page will be included in the “unknown”

section. In order to make the different categories as

distinguishable as possible, the color selection is

based on recommendations from the Colorbrewer

Web site (Colorbrewer, 2012). Of course, the user

can inspect a legend located in the visualization

space on the left hand side. By clicking on a

category in the legend view, it is possible to select

all Web pages from that group.

A common approach for bookmarking systems in

browsers is to store not only the URL and page title,

but also the favicon of a Web page. A favicon is a

small (commonly 16x16 pixels), squared icon that

identifies the domain of the Web page and usually

appears next to the address bar of a browser when a

Web page is loaded. To improve the chances that a

Web page is quickly recognized (Kaasten and

Greenberg, 2001) in the WebComets visualization,

the users have the option to additionally display

favicons over the pie charts, if these icons are

available. Due to the small size of the icons, the

users can still distinguish the sectors of the pie chart.

The size of the favicon is proportional to the one of

the circle. For small circles, the icons are also

reduced in size. To compensate for this, users can

hover the icon with the mouse cursor in order to

represent it in its original resolution.

When navigating the Web, users have multiple

options to reach a certain URL: type it in

themselves, click on a hyperlink, execute browser-

specific operations, etc. While some operations

suggest relationships between Web pages (e.g.,

hyperlink connections), others might represent a

breakpoint in the thought process of the user. Such a

case is usually given when users type in the URL

themselves. To better highlight this, Web pages that

are loaded after a manual input of the address have a

short vertical bar sticking out from under their glyph

to suggest a possible mental breakpoint, as indicated

in Figure 1.

Also, users may access a Web page by utilizing

the back and forward navigation buttons of the Web

browser. In many modern browsers, these buttons

allow the user to navigate backward and forward one

or multiple pages. These operations are useful

especially in cases where the users feel they have

reached an informational dead end. If a Web page

was loaded through a back or forward step, this is

captured and represented in the visualization through

small arrow-shaped elements on the left or right of

the corresponding pie chart (cf. Figure 3). The

numbers #b and #f inside the arrows highlight how

many back or forward operations were executed at

once in order to reach this Web page.

As not all these glyph elements are constantly

required, our tool users have the opportunity to

enable certain features like favicons and back-

forward arrows to be only included in the glyph on

demand. This allows them to maintain an overall

reduced level of complexity for the glyphs and only

access certain information when these could be

relevant.

4.3 Interaction

Besides the already presented features of the

approach, tool users have additional possibilities to

customize and interact with WebComets. The

current configuration can be stored inside an option

window and includes — but is not limited to — the

following: switching between comet tail and beam

representations, customizing min and max values for

the pie chart radii, enabling or disabling elements

like favicons or back-forward arrows, and selecting

the EBH attributes to be displayed.

4.3.1 View Transformations

A set of interaction metaphors complement the

visualization’s abilities by addressing topics like

flexibility, scalability or detail-on-demand.

The WebComets interface supports pan and

zoom operations, similar to modern interactive

maps. While the panning operation is self-evident,

there are two zooming approaches implemented in

the tool: one is a regular 2D zoom that allows tool

users to inspect the details in a certain area, while

the other is a 1D horizontal zoom along the timeline

that stretches the horizontal axis (Figure 4).

WebComets:ATab-OrientedApproachforBrowserHistoryVisualization

445

Figure 4: Temporal zoom (1D) along the horizontal time

axis: original zooming factor (left) and 30x horizontal

zoom (right).

Figure 5: Information box presenting details about the

selected Web pages. By moving the cursor over the list

elements, additional information is displayed for the

current Web page (in light-green).

It might occur that multiple glyphs are partially

or almost totally overlapping. Even if the tool user

has the possibility to execute timeline zoom

commands to compensate for this and clearly

separate the overlapping glyphs, this is a vital

scalability issue. To deal with this, WebComets

displays partially overlapping pie charts by

positioning the glyph of the Web page that has been

accessed later over the top of the previous one. To

further compensate, the tool user can move the

mouse pointer over a set of densely grouped circles.

By doing so, the glyph with the center closest to the

pointer will be moved to the foreground, over the

others. This focused glyph is additionally

complemented with graphical elements (e.g.,

favicon, if not enabled globally) and textual

information (e.g., Web page title). In cases where

the user wishes to focus his attention on the temporal

constraints and relationships, all glyphs can be

reduced to dots, minimizing the overlap of glyphs

and tails.

Furthermore, WebComets detects instances

where two or more glyphs are completely

overlapping and replaces these with the one that has

the largest radius. In order to suggest the presence of

other circles “underneath” it, the border thickness of

this circle is increased proportionally with the

number of hidden circles. Besides controlling the

depth order of the elements, moving the mouse

pointer over a glyph or its comet tail/beam opens a

tooltip that shows the title of the Web page and its

link as well as a set of page attributes customizable

by the user (e.g., title, URL, or pageviews).

Selecting a glyph can be done by mouse click, or

in order to select multiple glyphs, by pressing the

Shift-key plus mouse click. Besides this, tool users

can also use the legend to select all elements from a

category or search for certain patterns to highlight

glyphs that satisfy the search rule. Selected Web

pages are highlighted and an information box opens

in the upper-right corner of the visualization. The

information box contains the attributes of the

selected Web page, shown in Figure 5. This box can

also be customized in order to include all attributes

or a subset of them. In cases where multiple Web

pages are selected, the information box displays

initially only a list with the titles and URLs of the

selected pages. If the tool user moves the mouse

cursor over the list, the selected list element will be

expanded to present all available information of that

page. The background color of each list element

matches the category of the Web page. A click on

the link (or on the favicon if enabled) will open the

corresponding Web page in the tool user’s default

browser.

For each selected Web page, WebComets

searches the browser history to check for similar

Web pages. By default, only pages with the same

domain are considered to be similar, but more

complicated rules for interconnection can be

generated, e.g. pages containing a keyword, pages

that have an active time of similar length, etc. They

are highlighted by adding links between each

selected element and its counterpart. The curved

lines (Figure 6) are used for showing the presence of

similar Web pages to the selected ones, possibly in

areas of the visualization that are not currently

visible. Curves can be easily perceived as they

contrast with the overall orthogonal representation

of our approach. To avoid clutter, curves are drawn

in such a way that the probabilities of intersecting

curves and a curve intersecting a glyph are reduced.

More precisely, our tool computes a predefined set

of possible curves with different curvatures to

connect the two nodes and then utilizes a collision

IVAPP2013-InternationalConferenceonInformationVisualizationTheoryandApplications

446

detection algorithm to identify the curve that

intersects the least number of glyphs.

The tool user could find some parts of the

visualized history more interesting than others. Due

to this, but also in order to save screen space, he/she

has the possibility to collapse tabs or windows that

he/she is not interested in. For collapsing the

representation of a tab, the tool user has to click the

plus icon next to its spawning point; the same is

valid for windows. In cases when a tab is collapsed

that also has other tabs created by it, all the branches

of that tab will be compacted together with the

parent tab, and vice versa for expansion.

Nonetheless, if a glyph is selected and similar Web

pages were detected on any collapsed tab or

window, then the corresponding elements are still

visible by links and additionally small dots on the

tab lines, as shown in Figure 6.

Figure 6: Collapsed view: the presence of query-related

Web pages is highlighted on the collapsed tab lines by

small dots visible in the lower right part of the screenshot.

Connections between the selected pages and other glyphs

are highlighted through continuous curved lines.

Furthermore, in order to improve the scalability

of this tool and offer the analyst a better overview of

a large dataset, the collapsed view can be enabled

for all branches. In this mode, the vertical spacing is

compressed and the glyphs turn into small dots with

a colored halo, thus allowing the WebComets user to

inspect parallel browsing behavior in larger

histories.

4.3.2 Text and Motif Search

Highlighting different elements in a browser history

is closely coupled with searching for Web pages or

navigation patterns. Most Web browsers support a

text-based search of their records that limits their

ability to detect context information. This

functionality is also supported by WebComets. Tool

users can search for terms and keyword

combinations (e.g., apple+pc), strict phrases (by

using quotes, e.g., “apple pc”) or even exclude

words from their query (by using the minus sign,

e.g., apple–pc).

Besides the possibility to execute a text-based

search of the extended browser history’s

informational content, supporting the detection,

analysis and comparison process of temporal

patterns in the navigation graph requires a different

approach. Thus, WebComets offers a motif search

window (Figure 7) that allows users to define, store

and search for custom information and patterns of

navigation. Inspired by the building blocks concept

in (von Landesberger et al., 2009); (Reeder et al.,

2007), motifs in WebComets are predefined

subgraph structures that can be used to filter the

current history. Compared to other approaches, the

WebComets motif search has the advantages of

allowing logical combination of sought patterns (e.g.

find all node groups that satisfy motif A and do not

satisfy motif B), as well as the possibility to model a

large set of node and internode attributes.

These subgraphs can be generated in two ways:

either by mining substructures from the currently

opened history log files or by manually defining a

motif and its corresponding rules. In the first case,

the tool user would look through the visualized

browser histories and select any subset of glyphs

that could be at the core of a parallel browsing

behavior. When all relevant elements are selected,

the generated motif can be edited further in the motif

window to generalize or particularize the final

structural pattern. Contrary to this, in the second

approach the analyst would start generating a motif

by directly building it in the motif window, adding

node after node and customizing their attributes

based on his experience and assumptions.

As such, users analyzing an EBH can not only

look for topics of the visited pages, but also detect

navigation motifs. For example, the motif search

could detect that in 72% of the cases when an

Internet user accesses his/her e-mail account, he/she

also opens a Web page from the category “video” in

a new tab originating from his/her e-mail page. This

might suggest that he/she receives much e-mail with

links to video content. Thus, filtering the browser

history based on structural aspects can have many

applications, for example, detecting a Web page

where the tool user knows some attributes of the

originating site or investigating similar interest and

patterns of navigation between multiple users.

It is also possible to filter out numerical and

temporal values by giving exact numbers, suggesting

min or max thresholds, or defining intervals. These

rules are then incorporated in the motif and

displayed under the corresponding pages. In terms of

structure, complex motifs can be built by adding

WebComets:ATab-OrientedApproachforBrowserHistoryVisualization

447

multiple Web pages and highlighting existing

relationships between them, be it on the same

tab/window or on different navigation branches.

Once the motif specification is finished, the analyst

executes the search operation and the sum of all the

rules will be used for filtering the history. Finally,

the nodes that fit the query will be highlighted as

already described.

5 EVALUATION

A brief evaluation of the WebComets visualization

approach has been executed in order to capture any

advantages or disadvantages. The aim of the study

was to compare the performance and accuracy of

tool users when inspecting and comparing patterns

in multiple parallel browsing histories. For this

purpose, the participants would interact with the

same EBH log files by two different approaches: the

WebComets visualization and a list-based browser

history (the Firefox browser history was selected for

convenience).

The evaluation involved 20 participants with

experiences in visual analysis and/or knowledge

exploration as well as with extensive background in

using diverse Web browsers and accessing a variety

of online applications. All participants had prior

knowledge and at least some experience with list-

based browser histories. An initial step in the

evaluation process was to randomly divide the

participants into two groups and attribute a task to

them. Each member of the first group would have to

solve the task using the WebComets visualization,

while the members of the second group used a list-

based history. Next, the functionality of the two

tools was highlighted to each of the corresponding

groups. Note that the data contained in the browsing

histories were almost identical in content, except for

the fact that the list-based history was not able to

represent the additional fields generated and

included in the EBH.

The scenario involved the analysis of two

browsing sessions from different users. An initial

assumption was made that the users have

participated at an online conference call where they

suggested relevant Web sites to each other. This

collaborative browsing approach is frequently used

in cases where one party tries to highlight some

information to the other. A simple example for this

would be the collaboration between two remote

students that are preparing for an exam. As a result,

both users would access multiple similar or identical

Web pages in the same time interval. The test

persons had to determine if the initial assumption of

collaboration is supported by the browser histories,

and if so, what Web pages might have been involved

and in what time interval this collaboration took

place.

The results were evaluated by inspecting the total

Figure 7: The motif window that helps users construct, save and search for custom structures / motifs based on Web page

attributes and context information.

IVAPP2013-InternationalConferenceonInformationVisualizationTheoryandApplications

448

time each participant took to find a solution and the

time frame he/she reported as part of the conference

session. Figure 8 shows that the test persons

managed more quickly to find a solution with the

WebComets visualization than with the text-based

representation of the navigation history. To further

support this, the standard deviations have been

computed for the average solution times. The

importance of these values is further weighted by the

reported time frames for the assumed conference

session. Actually, the subjects that used the

WebComets tool have identified the correct time

frame in almost 100% of the cases, as only one

participant missed the presence of an additional Web

page in both history sessions around the same

moment in time. In contrast, the group using the list-

based browser history has reported a lower success

rate with only a 71% average overlap of the detected

collaboration time frame.

Additionally, the participants were given a post-

task questionnaire inquiring about the usability,

interaction and graphical aspects of the WebComets

visualization. Moreover, users had the opportunity to

express their opinions freely about the application

through a set of open questions. The outcome

suggested that users were overall satisfied with the

visual representations (e.g., “I found the comets had

useful information and were easy to understand”)

and interaction possibilities supplied by WebComets

(e.g., “I like the back and forward arrows. It’s a

better solution than repeating the Web page

instance” or “I can oversee all important information

in one view”), and that most participants would use

such a tool for analyzing parallel browsing behavior

or detecting and comparing browsing patterns.

Figure 8: Average time required by the subjects to find the

solution to the given task. The narrow black bars encode

the standard deviation.

6 CONCLUSIONS AND FUTURE

WORK

In this paper we presented WebComets, an

interactive visualization tool for tab-oriented,

parallel browser histories. After discussing related

work and functional requirements, we focused on

the individual components and interactive features

of our visualization tool. WebComets allows its

users to more efficiently search for patterns in

parallel browsing sessions as well as compare and

analyze the multi-tab browsing behavior of a group

of Internet users. An evaluation confirmed that our

approach has met the initial requirements, and our

users were able to quickly and efficiently gain

insight and find patterns in the history information

they were exploring.

As next steps of this research, we plan to make

WebComets widely accessible for analysts. While

the current version of this tool is focused around log

data, design and interaction features, in the future we

aim to adapt similarity search algorithms (e.g.,

automaton-based approaches) that could suggest

possible elements of interest to the tool users.

Furthermore, we plan to enhance our visualization

metaphor by adding a 1-dimensional, orthogonal

content-based operator for relative zooming along

the time axis in a certain area.

REFERENCES

Alexa: The Top Ranked Sites in Each Category,

http://www.alexa.com/top-sites/category, 2012.

Colorbrewer: Color advice for maps,

http://www.colorbrewer2.org, 2012.

Rescuetime, http://www.rescuetime.com, 2012.

Slife Labs Time Management Software,

http://www.slifeweb.com, 2012.

Adar, E., Teevan, J., Dumais, S.T. Large scale analysis of

web revisitation patterns. In Proceedings of the 26

Annual SIGCHI Conference on Human Factors in

Computing Systems, CHI ’08, pages 1197–1206, New

York, NY, USA, 2008. ACM.

Aula, A., Jhaveri, N., Kaki, M. Information search and

reaccess strategies of experienced web users. In

Proceedings of the 14th International Conference on

World Wide Web (WWW ’05), pages 583–592, 2005.

Card, S. K., Robertson, G.G., York, W. The Webbook and

the Web Forager: An Information Workspace for the

World Wide Web. In Proceedings of Human Factors

in Computing Systems (CHI 96), pages 111–117,

1996.

Cockburn, A., Greenberg, S., Jones, S., McKenzie, B.,

Moyle, M. Improving Web Page Revisitation:

Analysis, design and evaluation. In: IT & Society 1(3),

pages 159–183, 2003.

Cugini, J., Scholtz, J. VISVIP: 3D Visualization of Paths

through Websites. In: Proceedings of International

Workshop on Web-Based Information Visualization

(WebVis 99), Florence, Italy, pages 259–263, 1999.

WebComets:ATab-OrientedApproachforBrowserHistoryVisualization

449

Doemel, P. Webmap - A Graphical Hypertext Navigation

Tool. In Proceedings of the Second International

World Wide Web Conference, 1994.

Dou, W., Wang, X., Skau, D., Ribarsky, W., Zhou, M.X.

LeadLine: Interactive Visual Analysis of Text Data

through Event Identification and Exploration. IEEE

Conference on Visual Analytics Science and

Technology, 2012.

Eric, M., Ayers, E. Z., Stasko, J. T. Using Graphic History

in Browsing the World Wide Web. In: International

WWW Conference, pages 1–7, 1995.

Gandhi, R., Girish, K., Bederson, B. B., Shneiderman, B.

Domain Name Based Visualization of Web Histories

in a Zoomable User Interface. In: Proceedings of 11th

International Workshop on Database and Expert

Systems Applications (DEXA), pages 591–598, 2000.

Gross, E. F. Adolescent Internet use: What we expect,

what teens report. In: Applied Developmental

Psychology 25, pages 633–649, 2004.

Hightower, R. R., Ring, L. T., Helfman, J. I., Bederson, B.

B., Hollan, J. D. Graphical Multiscale Web Histories:

A Study of Padprints. In: UIST 98 Proceedings of the

11th annual ACM Symposium on User Interface

Software and Technology, pages 58–65, 1998.

Huang, J., White, R. W. Parallel Browsing Behavior on

the Web. In: Proceedings of the 21st ACM conference

on Hypertext and hypermedia (HT ’10), pages 13–18,

2010.

Kaasten, S., Greenberg, S. Integrating Back, History and

Bookmarks in Web Browsers. In: Extended Abstracts

of ACM Conference of Human Factors in Computing

Systems (CHI ’01), ACM Press, pages 1–2, 2000.

Kaasten, S., Greenberg, S., Edwards, C. How people

recognize previously seen web pages from titles, urls

and thumbnails. In: People and Computers XVI, pages

247–265, 2001.

Katz, J., Aspden, P. Motivations for and Barriers to

Internet usage: Results of a National Public Opinion

Survey. In: Internet Research: Electronic Networking

Applications and Policy 7, pages 170–188, 1997.

Kellar, M., Watters, C., Inkpen, K. M. An Exploration of

Web-based Monitoring: Implications for Design. In:

Proceedings of the SIGCHI Conference on Human

Factors in Computing Systems, CHI ’07, pages 377–

386, New York, NY, USA, 2007. ACM.

Kleek, M. V., Xu, C., Moore, B., Karger, D. R.

Eyebrowse: Real-time Web Activity Sharing and

Visualization. In: 28th ACM Conference on Human

Factors in Computing Systems (CHI’10), ACM, 2010.

Krstajic, M., Bertini, E., Keim, D. Cloudlines: Compact

Display of Event Episodes in Multiple Time-series.

Visualization and Computer Graphics, IEEE

Transactions on, 17(12):2432–2439, December 2011.

Miyata, Y., Norman, D. Psychological issues in support of

multiple activities. In: User Centered System Design,

D.A. Norman and S.W. Draper (Eds), pages 265–284,

1986.

Reeder, J., Reeder, J., Giegerich, R. Locomotif: From

Graphical Motif Description to RNA Motif Search in

Bioinformatics, In: Bioinformatics 23(13)

, pages 392 –

400, 2007.

Shen, S. T., Prior, S. D., Chen, K. M. A Solution to

Revisitation using Organic Bookmark Management.

In: Design, User Experience, and Usability. Theory,

Methods, Tools and Practice: First International

Conference, pages 46–52, 2011.

Spink, A., Park, M., Jansen, B. J., Pedersen, J.

Multitasking during web search sessions. In: Inf.

Process. Manage. 42, 1, pages 264–275, 2006.

Viermetz, M., Stolz, C., Gedov, V., Skubacz, M.

Relevance and impact of tabbed browsing behavior on

web usage mining. In: Proceedings of the 2006

IEEE/WIC/ACM International Conference on Web

Intelligence (WI ’06), pages 262–269, 2006.

von Landesberger, T., Rehner, R., Gorner, M., Schreck, T.

A system for interactive visual analysis of large graphs

using motifs in graph editing and aggregation. In:

Vision Modeling Visualization Workshop (VMV 2009),

2009.

Ware, C. Information visualization (second edition),

perception for design. Elsevier Inc., 2004.

Waser, J., Fuchs, R., Ribicic, H., Schindler, B., Blöschl,

G., Gröller, E. World Lines, In: Visualization and

Computer Graphics, IEEE Transactions on, vol. 16,

no. 6, pages 1458-1467, 2010.

Waterson, S., Hong, J. I., Sohn, T., Heer, J., Matthews, T.,

Landay, J. What did they do? Understanding

clickstreams with the webquilt visualization system. In

Advanced Visual Interfaces, pages 94–102, 2002.

Weinreich, H., Obendorf, H., Herder, E., Mayer, M. Off

the beaten tracks: exploring three aspects of web

navigation. In: Proceedings of the 15th international

conference on World Wide Web (WWW ’06), pages

133–142, 2006.

Yu, W., Ingalls, T. Trails-an interactive web history

visualization and tagging tool. In: HCII (10), pages

77–86, 2011.

IVAPP2013-InternationalConferenceonInformationVisualizationTheoryandApplications

450