utilizing structure based method of abstraction of detail. The ... - PDF document

utilizing structure based method of abst
utilizing structure based method of abstraction of detail. The major difference between FOXI and Bubble Tree is that the sub-categorizing is used not only as a method for abstraction but also automatically creates navigational points in the layout of the interface. The basic advantage of FOXI when compared to previously mentioned techniques is its ease of use, fairly straightforward concept that is easily understood by typical computer users, and stability of appearance during navigation even if the underlying structures are gradually changing. This enables effective exploitation of user’s spatial memory and fast learning capabilities. This paper describes the FOXI concept, its implementation into functional prototype, preliminary user testing results, discovered hierarchical tree or present novel navigational strategies that go beyond the traditional scrolling and panning either in 2D or 3D space. The visualization techniques are usually classified as overview+detail or focus+context approaches [8]. The principal difference between these two classes of interfaces is that the overview+detail techniques display the general overview of the rendered structure and the detail of the structure in focus in visibly separated areas, whereas the focus+context techniques use an integrated view utilizing special effects to enhance the perception of the area in focus. The FOXI technique is based upon assumptions similar to Bubble Tree [9] and can be classified as overview+detail approach. Just as the Bubble Tree, it is based on the property of simple to be searched using a search engine. In the end, the hierarchical structure needs not to be perceived as a contradictory approach to automated searching and may serve as its effective supplement. The major challenges in the area of visualization of hierarchical structures include making efficient use of the available display-area; prevent visual clutter and information overload, and the development of effective navigation approaches. Over the past years a lot of research work has been done in the field of effective display and interaction with hierarchically organized data. This research resulted in many different approaches like Fisheye Views [1], SemNet [2], Cone Trees [3], Tree Maps [4], Hyperbolic Browser [5] Disk Trees [6], Goldleaf [7] and others. These CR Categories: H.5.2 [User Interfaces] – Graphical user interface; E.2 [Data Storage Representations] – Composite structuresAdditional Keywords: Hierarchical structure visualization, zooming, limited display size 1INTRODUCTIONHierarchies are one of the most commonly used data structures. The principle of hierarchical organization is very well known and understood by most of the typical computer users. While a strong emphasis exists on the automatic retrieval of information through search engines that have already migrated to personal computers, the hierarchical structures still seem to keep their importance in situations where the context is not recorded along with the information it relates to or where the information structure is too FOXI - Hierarchical Structure Visualization Robert Chudý1Faculty of Information Technology Brno University of Technology Jaroslav Kadlec2Faculty of Information Technology Brno University of Technology ABSTRACTThis paper presents a novel approach in hierarchical structure visualization. The main goal of the presented approach is to achieve an ability to display infinite hierarchy size in a limited display area, maintaining high level of orientation and fast navigation in the resulting hierarchical structures. High level of orientation is achieved by specific hierarchy visualization techniques where each level of hierarchy is abstracted from its substructures and limited visualization area leads to high speed navigation engaging hierarchy substructure zooming. 2. Mouse moves – the structure of FOXI allows navigation with very little mouse movements. This enhances the ergonomic aspects of usage significantly. 3. Developing of movement routines – the users of the FOXI interface usually develop a routine of clicks and movements that allow intuitive browsing in the interface. Along with the stability of FOXI in situations, where the underlying structure is changing gradually and with the help of navigational points, this feature is a powerful rendering of user’s ability to employ spatial memory with the FOXI interface. This behavior is not forced and develops automatically, allowing users to browse at extremely high speeds. 5. Stability in gradually changing environment – FOXI proved to be stable in cases where the underlying structure got changed during the process of learning. However, the changes must ha

ve been gradual and not too complex. In
ve been gradual and not too complex. In these cases even some of the movement routines were easily adopted. 4. Unforced learning of the structure – the users reported an intuitive learning of the hierarchical structure rendered by FOXI. FOXI provides both general overview and the learning experience of details of the structure. 5. Orientation in unknown structures – after getting used to the radial layout even orientation in previously unknown structures became for most users easier. 4.2Problematic usability areas 1. Navigation points differentiation – the ability of FOXI to create navigational points is based on the differences in number of nodes among the sub-trees. A uniform distribution of nodes among the sub-trees may decrease the usability of this feature. However, the navigational points are not defined only by their structure but also by their position and corresponding description. This characteristic preserves user’s ability to navigate effectively with the help of his spatial memory. 2. Visual clutter – the ability of FOXI to abstract detail from the hierarchical structure is based on a complying distribution of complexity in the hierarchy. FOXI is primarily intended for shallow hierarchies, however with a proper method of abstraction it may be applied universally. If no abstraction method is available, FOXI may require interface enhancements to provide easy navigation. Figure 7. Interface enhancement option 3. Central node – FOXI, when compared to the tree browser, allows no jumps over nodes that accidentally happen to be very close to each other during browsing of the hierarchical structure. FOXI always displays a wide spread of nodes and allows no selection of nodes that will be displayed, whereas the tree browser may allow to jump from a node that is very deep in the hierarchy to a node high in the hierarchy because its sub-nodes have not been displayed. In this situation nodes from very different levels in the hierarchy can get very close and make the jump easy to perform; however, this feature is not controllable, happens mostly by accident and tends to diminish throughout the browsing actions. FOXI compensates for this characteristic with its fast back browsing and spatial memory features. 5CONCLUSIONThe usage of the interface proved that the FOXI interface is a very powerful navigation and structure-learning tool. The prototype was created to prove usability of the FOXI technique. Results of efficiency measurement indicate that the FOXI technique is not only an easy to use but also an efficient way to visualize and browse hierarchical structures. Future improvements of the prototype will include color coding to further enhance the appearance of navigational points in the structure, visual clutter reduction enhancements, operations and interface for adding and removing folders, bookmarks, and tracing of browsing history. ACKNOWLEDGEMENTThis work has been supported by research grant User Interface with Hierarchical Structures from FRVŠ MŠMT FR200/2005/G1. REFERENCES[1]Furnas G.W.: Generalized fisheye views, Proceedings of CHI’86 (Boston,MA),ACM,16-23. [2]Fairchild K.M.,Poltrock S.E.,Furnas G.W: SemNet:Three-dimensional graphic representation of large knowledge bases, Cognitive Science and its Application for Human-Computer Interface, Lawrence Erlbaum, NewJersey, 1988. [3]Robertson G.,Mackinlay J.,Card S.: ConeTrees: Animated 3D visualizations of hierarchical information, Proceedings of CHI’91 (NewOrleans,LA), ACM, 189-194. [4]Johnson B.,Shnedierman B.: Tree-maps: A space-filling approach to the visualization of hierarchical information, Visualization 1991, IEEE, 284-291. [5]Lamping J.,Rao R.: Laying out and visualizing large trees using a hyperbolic space, Proceedings of UIST’94, ACM, 13-14. [6]Chi E., Pitkow J., Mackinlay J., Pirolli P., Gossweiler R., Card S.: Visualizing the evolution of web ecologies, Proceedings of CHI’98, ACM, 400-407. [7]Faichney J., Gonzalez R.: Goldleaf hierarchical document browser, Proceedings of the 2nd Australasian conference on User interface, IEEE Computer Society, 2001 [8]Boardman R.: Bubble trees the visualization of hierarchical information structures, Conference on Human Factors in Computing Systems, ACM, 2000 [9]Cava R.A., Luzzardi P. R. G., Freitas C. M. D. S.: The Bifocal Tree: a Technique for the Visualization of Hierarchical Information Structures, 2002 233 FOXI - HIERARCHICAL STRUCTURE VISUALIZATION3. File window open – the prototype was developed as a functional application for general usage to show performance in real-life situations. This limited the options for effective and precise evaluation of the interface. Opening of file window helped to identify that t

he user has reached his desired destinat
he user has reached his desired destination folder. Double click used for this operation seemed to slow down interaction with the interface. Instead a middle mouse button click was used.4. Enhanced back – in cases where the direct jump to sub-folder was used, the right click on mouse provides jump back to the original folder skipping the folders on the way. In normal situations, left click in the middle of the base circle results in a move up in the rendered structure. 2.3Basic Visualization As mentioned before, the main task of the FOXI system is to visualize hierarchical structure in small area limited by circular shape. With exact knowledge of a hierarchy structure – its sub-structures – and radius of limiting circular shape a task of inscribed circles must be evaluated (fig. 6). Figure 6. Inscribed circles Knowing radius of the limiting circle R and number of inscribed circles N, representing number of sub-structures in the hierarchy, a function computing radius of inscribed circle can be evaluated. First a sector size in radians for each of the inscribed circles is computed (1) From knowledge of inscribed circle sector size a radius of inscribed circle can be evaluated (2) Radius R represents radius of the limiting circle,
. sector size in radians, and ri final radius of the inscribed circle. The radius of the virtual circle where the inscribed circles are visualized can be computed from (3) Further visualization of the next level of hierarchy is done similarly. Limiting radius is taken from the inscribed circle and the task of the new inscribed circles must be evaluated. The visualization of a hierarchy can be rather slow, especially in case of huge hierarchy with high structural complexity. In such cases, a limiting inscribed radius can be set to stop recursive visualizing at a level with radius that is too small to be rendered. The visualization process can benefit from utilization of this attribute as the physical visualization display has limited resolution and displaying of inscribed circles with too small radius has no positive effect on the visualization itself, because the displayed radius can be eventually smaller than 1 screen pixel. 2.4Zooming In and Zooming Out FOXI interface is using zooming technique for navigating in the hierarchy. Zooming involves limiting circle and one selected inscribed circle for zooming in or parent circle for zooming out. With these two circles and their positions and radii, a function of radius is computed to obtain appropriate position relative to the limiting circle. May l be a distance from the center of the limiting circle to the center of inscribed circle. Inscribed circle is resized by small delta radius
ûr (dependent on the zooming speed) and original radius before resize is saved to oldr. May radius of limiting circle be R and its position in zero coordinates (inscribed circle is positioned relatively to the limiting circle). Points initializing the final function are [oldr,l] and [R, 0] defining function of radius where distance from the beginning is decreasing with increasing radius. Final function computing position based on initial points is a linear function (4). Where y is the final distance from the zero and x is radius after resize for which the distance is applicable. Constants a,b,c are predefined as follows: With final distance after inscribed circle radius resized, the proper position can be computed. Similarly zooming out can be computed from current limiting circle and parent circle resizing radius down to the limiting circle radius. A problem arises with the
ûr increment. Too small increment sizes lead to slow zooming speed and are vastly slowing the whole interface operations. Too big increment leads to quick zooming making user feel little dizzy and disoriented with a question of “what happened?” and “where I am?”. The best results with increment coefficient have been reached with custom increment specifications as various users enjoyed various zooming speeds. 2.5Hierarchy caching Directory structure, as a source of hierarchy, works fine as the test subjects most usually know contents of their hard drive well and learning time to get familiar with the new interface has therefore been shorted to minimum. In case of artificial hierarchy, visualized by FOXI, users would have to learn and get familiar with this unknown hierarchy. This may extend the “get used to” time significantly. However, usage of directory structure on a hard drive brings several problems. The main problem is the drive speed and seeking speed. If the seeking speed of a hard drive would be about �����D�Ds

in1sin�� � Rri(2)
in1sin�� � Rri(2) idrRr�� (3) N�S�D� (1) (4) 0� ��cbyax(5) ����rroldalbcoldRbla� �� �� ��� �� 0231 FOXI - HIERARCHICAL STRUCTURE VISUALIZATIONn Systems, Computing Sciences and Software Engineering, 229–233. sub-circles are touching their neighbor sub-circles and the base circle. Figure 1. FOXI Layout1email: chudy@fit.vutbr.cz 2email: kadlecj@fit.vutbr.cz 229© 2006 Springer. T. Sobh and K. Elleithy (eds.), Advances advantages, and disadvantages with corresponding enhancements and future work related to FOXI development. 1.1FOXI The primary objective for FOXI was to allow rendering of hierarchical structures with unlimited size in limited display area. This ability makes FOXI ideal for navigation over structures growing in space, like virtual cities and data landscapes. FOXI is usable in cases where the displayed hierarchies reach a moderate complexity. For dealing with highly complex hierarchies, enhancements need to be applied both over the hierarchy and the interface utilizing FOXI technique. This situation is described under the Usability review section of this paper. The layout of the resulting interface structure is created by This technique allows intuitive recognition of spatial relations among displayed sub-trees. Despite the usage of the same building elements, different number of nodes in the sub-trees results in automatic creation of easily recognizable navigational points. These navigational points are simple to read and bring stability to the appearance of the structure even if the underlying hierarchy is gradually changing. Figure 2. Navigational points Increased density of nodes in the displayed sub-trees results in creation of higher number of visual signs in the area. FOXI allows free rotation of the sub-trees around the axis of the base circle what gives variability to possible setups of navigational strategies. FOXI technique tends to preserve the most stable inner distribution of visual elements to make the best use of user’s spatial memory and to allow reading of the context of the rendered structures thus allowing not only their perception but also understanding. 2IMPLEMENTATIONThe prototype of FOXI described in this paper is a zooming interface that works with the hierarchy provided by the file-system in common computer system. In this case FOXI displays the folder structure only to limit the number of displayed elements and to avoid visual clutter as much as possible without modifying the construction of the hierarchy it displays. To display the files a separate window is used. 2.1Basic layout FOXI interface visuals implemented in the prototype include: 1. Display area with navigational points 2. Textual description of folders on the current level – these descriptions work as supplements to navigational points. In case of properly structured and simple hierarchies it is possible to make them display only temporarily and over the display area. 3. Back button allowing moving one step up in the hierarchy – the placement of the button in the middle of the structure is strategic and allows very fast movement up in the hierarchy. The positioning in the same place throughout the whole navigation process requires no movement of mouse when moving up the hierarchical structure. Figure 3. Basic layout with back button 2.2Enhanced layout and interactions FOXI was originally developed as navigational gizmo concept for a spatial interface controlled with hand-gestures. The prototype described in this paper is controlled by standard 3-button mouse;therefore, minor changes in visuals of the interface and control interactions had to be implemented: 1. On mouse-over highlight – the highlighting of current selectable circle allows user to easily identify his current options for movement in the structure. This feature is especially usable when jumping directly to a sub-folder. Figure 4. On mouse over highlight 2. Direct jump – allows user to jump directly into a sub-folder skipping a number of folders on the way. With this feature active, both On mouse-over highlight and name display of the current selectable folder are rendered. Figure 5. Direct jump highlight with folder name 230 CHUDÝ AND KADLEC10 ms, reading of a hundred directories would last over 1 second. Typical hard drive containing several hundreds of thousands of directories would take a very long time to load. Even visualization of current directory with 50 dire

ctories visible in depth would get stuck
ctories visible in depth would get stuck for almost a half of a second, what is very inconvenient. A caching system has been proposed to solve this loading problem raised from the hard drive speeds. External cache file has been created to hold hierarchy data. Main benefit of external file is that FOXI can be brought to virtually any environment with its own structure representing menu with actions, database with images and so on. Test users got, thanks to this external cache generation, very fast tool for browsing contents of their hard drives. FFICIENCY EASUREMENTAs stated before, with FOXI prototype users got a tool for browsing the contents of their hard drives. Several approaches were proposed to test the efficiency of the proposed FOXI structure. Approach with predefined hierarchy structure and set of tasks to do seemed to be a good idea as it would measure browsing speed in distinct tasks. However, this approach required users to learn to control not only the user interface but also to learn the structure of the test hierarchy. With too simple hierarchy structure there would be only small differences between tree browsing and FOXI browsing so another approach has been proposed. Users working with well known hierarchies would have to learn the new user interface only and the total learning time would be much shorter. Using FOXI with their own hierarchies with no specific tasks to read and perform was for most of the users very engaging. In this case the goal of measurement was to acquire all possible data from user’s browsing actions. The problem arose because of the need to distinguish the moments when the user started to browse and he stopped to browse in the hierarchy. The user typically stops browsing when he finds certain information and displays it. In this case a window is opened. Opening of the window takes focus from the testing and measuring application. This action is considered stopped browsing. Similarly, when the user opens FOXI application, he starts browsing or navigating through the structure for some kind of information. This moment of gained focus is considered the start of browsing. Between these two moments, application measures number of levels that the user has browsed and the number of clicks she made to reach her goal. The time measured between focus gain and focus loss indicates how long was the user browsing through the structure. However,this time is not precise because the user could spend some time to open the application and the time to the first click could be very long. That is why a time from the first click is also measured to eliminate these problematic parts of browsing time. As no other possibility exists how to compare the FOXI approach, a secondary application has been developed. This application was similar to the tree browser, featuring similar set of functionalities like FOXI and measuring the same actions like time of browsing, number of clicks and number of browsed levels. These values can be compared, giving approximate level of differences between these two approaches. Table 1. Table with FOXI measured data by novice user Levels 2 34Clicks 2 34Total Time (ms) 3355 6022 8805 Time from first click (ms) 3345 5835 8091 Table 2. Table with FOXI measured data by skilled user Levels 2 34Clicks 1 22Total Time (ms) 1823 2364 2828 Time from first click (ms) 1012 1512 1802 Table 3. Table with Tree measured data by novice user Levels 2 3 4 Clicks 2 34 Total Time (ms) 3145 4956 6732 Time from first click (ms) 3124 4812 6830 Table 4. Table with Tree measured data by skilled user Levels 2 3 4 Clicks 2 3 4 Total Time (ms) 2183 2933 3835 Time from first click (ms) 2183 2489 2864 3.1Efficiency measurement results The efficiency measurement was focused on the speed of browsing only. The results indicate that the FOXI technique is slightly slower than the common tree browser when used by novice users. However the learning curve for FOXI interface is very steep, what makes the initial speed unimportant. In case of skilled users, FOXI usually performs better than the tree browser. In the best observed situations, FOXI performs more than twice as good as the tree browser. The slightly worse results between the total time and time from first click, even if in favor of the FOXI, indicate a slower speed of user orientation in the radial layout of the interface. After the initial click, the spatial memory is on work and browsing times improve significantly. SABILITY REVIEW4.1Positive usability features 1. Ease of use – if the direct jump feature is not used, browsing of the FOXI structure requires no precise aiming with mouse. In most cases, the circles representing nodes are large enough to be hit with low accuracy of aiming when compared to the tree brow