Reflecting User Perspective in Document Spaces
Eric Gould Bear
InterCHI 1993 • April 24-29, 1993
As the ease of accessing and generating large quantities of information increases, people’s ability to navigate through that information and maintain personal perspective decreases . This paper describes an interface element, the Relativity Controller, that enables users to specify what is important to them and modify the portion of their perceptual space that
information takes up, using a variation on fisheye view techniques . This process is described as a generalized tool
for annotating documents and for controlling the balance
between detail and context in representations of document
contents. Peripheral portions of documents are condensed so
that salient segments can be expanded and whole document
contexts maintained. It will be shown here in its application to
User Interface, Fisheye Views, Personal
Perspective, Annotation, Information Retrieval, Video Editing,
While fisheye views are common in human cognitive processes, such as perception and memory, and have been demonstrated as useful in computer interfaces , they are rarely used in common computer applications. This may be due to a lack, thus far, of suitable interfaces for personalization of documents.
The need for this type of control was expressed most recently by Mills, Cohen & Wong . They enabled users to magnify details and maintain displays of context by creating a hierarchy of video data. Users successively extract increasingly detailed segments from larger contexts. However, this method required numerous user actions and filled the electronic desktop with several windows of video, thus bringing up questions of efficiency and cognitive overload.
The Relativity Controller addresses these concerns by maintaining a single window to the data and giving users the ability to adjust the degree of fisheye magnification in a single step — the user always has visual and navigable access to the whole document via the scroll bar.
In order to enable users to navigate among different perspectives from the scrollbar, the scrollbar itself must maintain a fisheye representation consistent with that of the document. Chimera  demonstrated scrollbar satellites, called Value Bars, that graphically indicate relative values of data attributes independent of the actual data’s representation in a window. This is accomplished by scaling proportions of attribute indicators according to their values in a bar parallel to the scrollbar. The Relativity Controller leverages off this idea, but instead uses a single bar to perform both the functions of scrollbar and Value Bar. Indicating the scope of salient segments in the scrollbar allows users to simultaneously navigate through the document and transform its represented meter — from regular intervals to intervals that scale according to salience.
APPLICATION TO VIDEO DATA
In this implementation of the Relativity Controller, users are
provided with a means of annotating and modifying perspective
on digital video stored in the QuickTime movie format on Macintosh computers. The standard Apple QuickTime
controller, usually attached to the bottom of a movie's window,
is replaced by the new controller shown in Figure 1. As in the
standard controller, there is a 'play/pause' button and a scrollbar
representing the duration of the entire movie. People can use
this new controller to view movies using established methods.
Figure 1. Adjusting perspective on video data. In this figure,
the user has marked three salient segments of video. Notice
how these marks in the scrollbar get wider, as the spacing
between them condenses and darkens, when the mouse is
moved up. The cursor, to the left, changes to reflect the size of
the marks in relation to the length of the whole video.
To suggest to the user that they can mark segments of video, the cursor changes into a stamp tool when it is in the bounds of the movie's window. While the video is playing, pressing and releasing the mouse button marks a segment in time corresponding to the duration of time the button is held down. Otherwise, if the movie is paused when the mouse is clicked, a single frame is marked. Small colored marks appear in the scrollbar, representing the scope of the marked time in the movie. Users can easily review these segments by moving the scroller thumb to a mark and clicking on the 'play segment' button, at the far right of the controller.
Users can now focus their attention on the salient segments of the movie by shrinking the representations of unmarked or insignificant segments. Clicking on the 'scale perspective' button, at the far left of the controller, changes the cursor into a slider that can be moved vertically. Doing so affects the relationship between the marked and unmarked segments. Moving the cursor upwards causes the unmarked segments to condense, allowing the marked segments to expand and fill the remaining scrollbar space. Condensing time is reinforced by a darkening of the corresponding scroll areas; expanded time is lightened. This contrast makes the extent of fisheye magnification obvious to the user.
Since video occurs in time, magnification is also applied to the playing video — speeding up through condensed segments in proportion to their spatial density in the scrollbar. As a result, users maintain a single model in space and time of the degree of magnification into the salient portions of the document.
Initial user studies have indicated that the relativity controller is an intuitive method of establishing and representing personal perspectives in video documents. Video editors found it useful as a pre-editing and archiving tool. Others saw it as potentially enhancing interpersonal communication: sharing annotated and scaled video that retains its original context might facilitate a better understanding of unique points of view.
Issues that need to be further addressed include methods for modifying the scope of marks, a more obvious means of initiating perspective change, and ways to represent layers of collaboration in annotation. In addition, I plan to apply the controller to other data types. Figure 2 shows a rough sketch of how text could be presented as a pre-scaled abstract in search systems . People could scan and retrieve information at personally relevant levels of magnification, thus making the online inclusion of complete works a viable option.
The Relativity Controller is an important development that has practical application to the navigation and visualization of various types of information, enabling people to construct data spaces that are reflective of their personal perspectives.
Figure 2. Sketch of Relativity Controller for text. The user
expands a pre-scaled text abstract by dragging the mouse to
the right. The windows to the right show the resulting succession
of increasingly detailed segments of text coming into
view. Notice that the size of the marks in the scrollbar show
relative proportions of marked and unmarked text being
Thanks to Kenny Miller, David Brick and Jonathan Cohen for inspiration; Red Burns for encouragement and support; Lili Cheng for graphic design assistance; Meg Fry, Marc Lenburg and Richard Gould for helping me articulate concepts; The Human Interface Group at Apple for invaluable criticism of my early abstract attempts to warp time; and Janna Buckmaster.
1. Mackinlay, J., Robertson, G., and Card, S. The Perspective Wall: Detail and Context Smoothly Integrated. In Proceedings ACM CHI’91 Human Factors in Computing Systems Conference (New Orleans, LA, April 27-May 2, 1991). ACM, New York, pp.173-179.
2. Furnas, G. Generalized Fisheye Views. In Proceedings ACM CHI’86 Human Factors in Computing Systems Conference (Boston, MA, April 13-17, 1986). ACM, New York, pp.16-23.
3. Mills, M., Cohen, J., and Wong, Y.Y. A Magnifier Tool for Video Data. In Proceedings ACM CHI’92 Human Factors in Computing Systems Conference (Monterey, CA, May 3-7, 1992). ACM, New York, pp.93-98.
4. Chimera, R. Value bars: An Information Visualization and Navigation Tool for Multi-attribute Listings. In Proceedings ACM CHI’92 Human Factors in Computing Systems Conference (Monterey, CA, May 3-7, 1992). ACM, New York, pp.293-294.