,
‘3 Eser Kandogan and Ben Shneiderman *
l
l
l
l

The cognitive overhead problem is’more related tdtheuser’s -, :
browsing strategy. It is about how users seek-a-balance b - ’ -
tween the gains of added knowledge and the;lossy &om : : ;
increased istraction by following a link. Marchro~mm and ” ’
Shneiderman [ 161 argue that in browsing, goals,are not-,well
defined and change dynamically as new iirformatrcn~ IS en- 2. ;a
countered by the user. ,r,-: y 2 J , ,-.
1 -- ’ <.:*.: _._. *,ht, ,‘;!:-,*i’;lj,.
Cruz [lo] observed that current web browsers give little flexz- . i .,.. ‘6.. i :_ .a ‘
ibility to users both in filtering out unwanted informatton and ‘1
in the specification of the spatial and temppral lavo&~ This, I,
;
‘..? L. 92 i’i: .,
‘-j ? ” :’ 1
l High page recurrence rate: 58% of the pages revisited
l Continued grov&&JRL vocabulary
l High recency of revisits
l Frequent visits c@y to very few pages: Personal and 01
nization pages, search engines, etc.
l Browsein smalk clusters of pages
l Short sequences .of repeated URL paths
-EL&TIC tiir;DbWS WEB BROWSER
limits user’s browsing and organization!r?~pabilities.“~“ rHitiraichical Page Oiganization The Elastic Windows browser is a multi-window browser,
where pages are organized hierarchically [14,15]. Hierarchi-
cal organization of pages allows the user to see the context,
while exploring further details lower in the hierarchy. Al-
though syntactic information structure on the WWWjis an
arbitrary graph, presenting the information in a hierarchy can
help users in their information seeking activities. While hits
archical organization might8facilitate’navigition, t can alsd
giveusers a sense of location in the information structure. a 9 r ‘,
The fundamental mechanism for organization iscomposition.
However, Halasz [ 131 argues that the hypertext model acks
a composition mechanism, i.e. a way of representing.yd
dealing -with groups ofnodes;$nd lmks as unique?mttltres
separate ,fromtheircomponents.~ :I) - : .‘- .- <‘j,-, id-c
‘L,, -, . !~i. !.f I, ,!‘.‘! :- - ri,,. ’ * *,Ir,,” _ / _’ !/ ; , rj :: .‘I-
‘Rosenberg [ 191 pointed out that current WWW browsers pro,
vide*a single window,on the document, and when user&click
to follow, a:$&, the new documentis opened in place; r&
placing the ‘former. Some systems allow’ another window
to be opened for the new document. Browsers simply rely
.on thewindow manager to organize these open docum&ts.
However, current’.ivindciw managers fail.,to provide ahoiga:
nization whichreflects the semantic relationship that-exists
among documents browsed on the WWWJ :I ,*ri’ : ,_... . -; J
:I- ‘--,
LESs.ONS 5pOfA~lJSER &DIES”
. I. r;, *T;ir?r,9:.‘” “, . ‘-:_
’ ’ . .:,I; II&:: ..) i(
Although several.,studies provide,deimographicalinformation
on users and web-sites, our interest is on studies that examine
users’ navigational ccess patterns. .Studies Ydone on-users?
navigationalstrategies by Catledgeand Pitkow [6] and by
Tausche’r and Greenberg [2 I] am,particularly-interesting ‘i
thatthey are,done‘i’n open systems forlo@ dtiiations:’ . :
_ - - -
I ;-L-J, I.1 .,:; ._^. - ‘, I::‘, -,f:?::, -:.,.* .v- ./f
Catl&lge’&d Pitkow captured client-side’user-events of the
XMosaic browser from a uonulationof 107 users in the Geori
gia +stitutepf Technologjr’~College of Comput$g for a three
week period. . : ‘~ I. I(..’ g ,?’ . :;
They characterized users’ navigation.strategies according to
the average frequency of followi;i~‘~~“~~t~“vtith a certain
depth. Average frequency was fotiird to be linearly dependent
on the, path length with a slope .ofTQ,.24. Users’ browsing,
strategies are classified q serqxJip_itou~ browser,) general
purpose browse?;:apdsb~ed on users’ average slope.
1; --; ‘iF*
Besides this classrficatioz of the navigation&-ategies:‘they
also &se&d: s.ii * !. ‘T~,;G cr.’ r ;,-‘l j - ,$I s
-,,l-,;<“-J :.,,,:+ .$ ?
i Tendency ,to‘browse’ ina small area “’ ’ ’ - ‘: *”
I
;:’ Fre@enf I&! ‘of back&ickingi .B&k button usage-4 lo& “,-
l Shallow browsing: -Rarely ‘more than two layers _’
:
l Infrequent save, print, hotlist addition and retrieval ’
Tauscher .and..Greenberg. analyzed 6 weeks of detailed us-
age data from 23 users with at least a-year experience of
browsing on an instrumented XMosaic browser. They ana-
lyzed recurrence of page&its, growth of URL vocabulary,
visit frequency as a function of,distance, ,f$quency of,URL
accesses according to page types,
peated sequences:’ ”
locality, and length of re-
”
,In’summary; their observations are:
*go-
$gure 1,: shows a user browsing ,the Human-Comp&r In-
tern&$ Lab (HC,$) web pages. Pages are~.hierarchically
organized with the top level HCIL main page placed on the
@kl.On the I+&~ four pages (Lab Rescription,, P$ncipal
geqb&s, Students, CoJlabotitqrs, and,Research Project De-
scr@tidn‘pages) are opened as a group in the Ahout HCI~
window.-,Furthermore, fiveproject pages are opened in tho _I.__ ,~
research Project Description page. The hierarchy is created
9 aresult of user actions., not prepared@ advance. ‘I.‘,
C&r&7 n.&vp&~ ‘In the Elastic Windowlbrowser, z new
page can be opened by clicking on the link. The window for
the new page is opened “inside” the window of the parent
page; placed on the right, using half of ,@e space (Figure 2.a).
Thus, when a link is followed, the context is preserved on the
left, while the detail on a link is being examined on the right,
By selecting adifferent link, the user might either eplace tho
last selected link(s) or add the new link to the existing pages
sharing the space. , ’ 2 .’
Multi&links& be opened by the select operation with the
lef&&e~button followed-by:a right ‘mouse button click,
All selected .link&re opened side by side, placed !within
their parents window on the right (Figure 2.b). ,,Alternatively,
vertical and tiled placement styles can be selected from the
window menu. Links in a region can be selected by drawing
a rectangle.on the.links;.with t e left mouse button. ‘Non-
contiguous lii can be added to the selection by pressing tho
control key with the left mouse pressed. I ’ /*, (,)
The conventionalopen-and-replace strategy is still available
in theElastic Wihdowsbrowser by clicking the right m&so
button with the Control key pressed. This way users can skip
uninteresting intermediary pages-when following a number
of links to the desired ,information. 111 i l,’ / .“) I I ‘II ‘I
Multi-/eve/$&k ,Whknbrowsing a large information strut-
ture; users r&ight.w& to keep;multiple vels of contexts on
the screen at the same time. This might help users in their
navigational strategy bygreatly reducing the need to back up,
and also lessens apossible disorientation.
; 1
I 170