Using Space Effectively: 2D II
From CS294-10 Visualization Sp11
Lecture on Mar 9, 2011
- A general cartographic labeling algorithm, Edmondson, Marks & Shieber (pdf)
- A survey of automated layout techniques for information presentations, Lok and Feiner (pdf)
- Dynamic space management for user interfaces, Bell & Feiner (pdf)
- Rendering effective routemaps, Agrawala & Stolte (pdf)
- Artistic resizing, Dragicevic et al. (pdf)
- Adaptive grid-based document layout, Jacobs et al. (acm)
- Map labeling bibliography
Manas Mittal - Mar 09, 2011 01:31:11 pm
Reading the Feiner paper, I am intrigued to think about automated layout manager in apps like Flipboard. It has beautiful generated automatic layouts. Does anyone know how that works?
Brandon Liu - Mar 10, 2011 01:58:18 pm
After looking at the Edmonson/Marks/Sheiber paper, it would be criminal not to mention this series of blog posts on Google Maps layout : http://www.41latitude.com/post/2072504768/google-maps-label-readability
Saung Li - Mar 10, 2011 03:35:51 pm
I'm not a big fan of distorting maps in general, as they tend to draw my attention away from what the visualization is trying to show. Instead of first thinking about the purpose of the image, I think about how the map is distorted and how it is different from what it should be. I'm not sure if other people feel the same way, but since I learned how maps really look like, such as the map of the United States, if someone distorts that then the first thing that comes to mind is "they drew this map wrong." As such, I think distortions impair my ability to decipher what the visualization is to show, as I first have to tell myself that the map is distorted for some purpose. I felt that most of the distorted maps of the US shown in lecture had this problem, and I'm wondering if there is a better way to do this. Some distortions work well, though, such as the underground train maps, as people don't need to know the exact roads and mainly just want to know which stops to get on and off.
Michael Hsueh - Mar 10, 2011 07:36:31 pm
The Lok and Feiner paper provided a nice survey of the field of automatic UI layout generation. I am by no means a web designer, so I could be wrong, but CSS + HTML seems like a half abstract, half spatial layout specification framework. The bit on learning and evaluation techniques was interesting and illustrated the important challenges facing those tasks. "What is the right way to 'benchmark' a given layout?" is perhaps the most critical question underlying discussions of how learning or abstract-to-spatial translation methods should be designed. The paper's idea of an adaptive or even dynamic layout that is calibrated to individual users is one interesting approach to this question.
Marks and Shieber's method of cartographic label generation is detailed and makes a lot of sense. I liked the bottom up approach to a large problem and the application of simulated annealing to find the globally optimal set of labels. Basing the final evaluation function on a weighted sum of metrics derived from intelligently designed heuristics seems like it could lend itself nicely to machine learning techniques. The framework presented is important in that it allows adaptation to different layout benchmarking models by simply tweaking heuristics and metrics. The paper on automatic layout generation does not get into the details of abstract-to-spatial specification translation... but I can imagine that a similar techniques involving simulated annealing etc. can be (and probably is) used to choose optimal layouts from among those that satisfy specification constraints.
Dan - Mar 10, 2011 08:50:27 pm
Survey of Automated Layout Techniques for Information Presentations: This paper was very interesting! First of all, layout is everything. We experience and read layouts everyday probably on the order of 100s of times. The layout clearly has a lot to do with how much information can be transmitted to the human visual system. The simply layout seems a bit simple (pun intended), but a great deal of improvement occurs when you add some spatial constraints, as shown in figure two of the paper. Overall this paper was good and intriguing for a potential research project!
Dynamic Space Management for User Interfaces: A very novel approach to user interface screen real-estate and using windows in a new way. I like the system of have the classifications full and empty. I also think that using axis-aligned rectangles makes for a good approach for rectangular windows and 2D GUIs. I wonder how they classify what is a good layout, I'm guessing that they can use some psychological tests to measure the information bandwidth, but what about aethetics?
Sally Ahn - Mar 11, 2011 02:01:51 am
The Dynamic Space Management paper made me think of the Expose's "All windows" functionality on the Mac. It has become a feature I use frequently on daily basis, but at the moment, it simply lays out all the windows in a grid. I know I have often gotten lost in a sea of open windows, and it would be great if some of the automatic layout techniques could be applied to allow users to navigate easily from among multiple windows. It seems that the new Mission Control on Lion implements something like this.
Karl He - Mar 11, 2011 07:25:54 pm
The Dynamic Space Management paper presents an interesting idea, that I see as most applicable for a PC window manager. I do however feel that some of its ideas are misguided. For example, even if I wanted to take as much room as I could for my browser window, by no means would I want it to be a long rectangular shape, I'd much rather have a more squarish shape. In addition, the adjustment of windows when they are dragged might be more disorienting than beneficial. I also don't feel that preventing things from overlapping is necessary. All that is needed is sufficient visibility of background windows such that the user can discern what the window is.
In general, my main complaint is that the dynamic reorienting of elements may be more disorienting than beneficial. If a display was displaced, I may have trouble relocating it. The part that I find would work is the dragging behavior. Slight displacement during this would still be trackable by a user, and would result in more satisfying placement.