Download Analysis of SequoiaView and Grokker: Two Hierarchical Visualization Programs and more Papers Computer Science in PDF only on Docsity!
Nicholas Diakopoulos Information Visualization CS HW # February, 24 2004
The goal of this paper is to analyze a couple of hierarchical visualization programs: SequoiaView and Grokker. Evaluation will proceed along several lines: relative strengths and weaknesses, suitability to certain types of tasks, and any user interface deficiencies that may be present. In general we should first consider the types of tasks that are encountered in a hierarchical information visualization and browsing context. Wehrend and Lewis define several tasks important in information visualization such as: identify, locate, distinguish, categorize, cluster, compare, correlate etc. Of these, only a few are truly relevant to the evaluation of these programs. Clustering, categorization, and location seem to be primary tasks that one might engage in when browsing a hierarchical information structure. For instance, one might want to visualize clusters of certain types within the hierarchy or to locate a given file or file type based on its category. Navigation and browsing are other important activities involved in hierarchical visualization.
SequoiaView SequoiaView is based on the treemap visualization technique and is specifically oriented toward the domain of file system visualization. Because it is quite specialized to filesystems it does a fairly good job overall of providing some sense of the structure and organization of the files on a drive. Since the area of each element in the tree map is proportional to the size of a given directory of file, a clear strength of the visualization is that it gives a very quick overview of where and how space is being utilized on a drive. An application of the visualization technique to another form of data would be interesting to see, but unfortunately the tool is not that general. A weakness may be that in the absence of a simple metaphor (such as box size to file size), the interpretation of other information hierarchies may not be so intuitive using the treemap. That is to say, some element of the data needs to naturally and intuitively map to size in the display. A weakness of the treemap in general is that it can produce very long thin rectangles which are difficult to navigate. This is somewhat alleviated by the squarified tree map.
SequoiaView was found to be fairly good for seeing clusters of filetypes using the color palette functionality. One can very quickly see where there is a concentration of image files for example, by selecting the image color palette and then looking for the corresponding colors among all of the other grayed out files. Again, this coloring helps for categorization of file types and for locating file types (other tasks that we defined). See the following figure which shows the image filetype palette in use. In general, the flat color scheme was disregarded since it was so garish and made navigation much more difficult than when the cushioning was turned on.
Filtering functionality in SequoiaView is somewhat weak insofar as it is not dynamic. One can filter on file size or date which is of course useful when looking for recent files or for files taking up a lot of space. However, dynamic query capability would bring the tool to a whole new level of usability. When someone is browsing for a certain file from a given time period, he may only have a fuzzy idea of what the date is and thus interactive feedback on the date filter would be quite helpful. The implementation of interactive feedback may not be feasible for large file systems however. I found the different types of highlighting and selection interesting. The option which would highlight not only the file, but also its entire path up the hierarchy by using a graduated color scheme was helpful. It lends context to the file and where it is situated in the larger directory structure. I feel that this highlighting mechanism gives more insight into the hierarchical structure than the other highlighting options. Additionally, the tool tips in general provide an
Grokker Grokker is a more general hierarchical information visualization tool which can operate not only on file system data, but also on amazon.com product data, and general web search data. Like SequoiaView there are several color options which make the tool capable of categorizing and to some extent also clustering based on color responses to dynamic queries.. The key strength of Grokker is that it supports smooth zooming and browsing of the information hierarchy. This allows the user to maintain some sense of context for the data that he is browsing. Zooming in and out is very easily accomplished with a single click. I found it particularly helpful to be able to see one level up in the hierarchy so that you can easily navigate back up a level. The ability to control zoom speed is another nice feature since for human factors issues, some people might prefer a slower zoom than others. Not that I ever got lost, but if I did, the home button is a nice feature because it jumps you back up to the top level view of the hierarchy. I found the explorer window functionality of the system to be a strength as well since whatever is in focus in the visualization window can also be viewed in the explorer window based on whatever kind of data it is (e.g. webpage, image etc.). Thus the system allows for easy browsing to a given point in the file system and then the ability to change the contents of a folder in the explorer view, which is something that SequoiaView doesn t support. A drawback of the circular style visualization is that there can be a lot of wasted screen space at the center of the circle. If there are a lot of categories, then the circles are small and the circumference of the root circle must accommodate all of them. This can be changed by having everything render as squares, though this is not only less aesthetic, but somehow on a very low visual level, makes it more difficult to see the differences in shape size between levels of the hierarchy. That is to say, that the hierarchy is more apparent and appealing using the circle rendering. I found that there needs to be a better key for the color coding. It s easy to see the relative difference between colors, but not to see how a color maps directly to a value such as filesize. The filtering ability of the system is fairly good. Specifically, dynamic query functionality allows for interactive visualization strategies which provide a clear advantage over the static queries of SequoiaView. The option to either gray out a category or make it disappear when filtered out is a nice feature too. I found just graying out things in the course of filtering was better since it kept the overall layout the same and thus there weren t any abrupt changes in the layout and context of the visualization which could be distracting. I found the opens with
filter especially novel since I could imagine a user task involving knowing the program used to create a file, but not remembering where it was saved. All in all, I found Grokker to be a very effective browsing tool. It combines an intuitive zooming interface, detailed tool tips, and even contextualized/ rendered images within the visualization to provide multiple layers of information in the course of browsing. For instance, when searching for Depeche Mode on Amazon I could see the interrelationship of all Depeche mode merchandise such as CDs, DVDs, and books which made it consequently easier to browse. The rendered icons were very helpful in this case since I could recognize certain albums based on cover artwork. Despite the easy interface, I m still not convinced that I should give up my regular web browser though! Perhaps a weakness of the system is that it attempts for form categories in web search information which can be inaccurate. I grokked Diakopoulos and at first was impressed that it had created separate categories for me, my brother, and my father. However, upon closer inspection many of the classifications within these categories were in fact wrong, which is obviously misleading. At the same time, the filtering was again impressive because I could filter based on web domain and see that all the references to Diakopoulos coming from Germany were for me (apparently I m the only Diakopoulos on the internet in Germany!). In summary, I found Grokker to be a much more general, polished, and functional program than SequoiaView, which is logical given that Grokker actually costs money to buy. Though the interface was a little bit rougher in SequoiaView, its specialization on filesystem visualization makes it a useful tool for tasks in that domain. Grokker however, supports more user tasks insofar as it can also be used to visualize amazon.com product information and general web search results. Overall it does a better job at browsing and filtering than SequoiaView.
SequoiaView was found to be fairly good for seeing clusters of filetypes using the color palette functionality. One can very quickly see where there is a concentration of image files for example, by selecting the image color palette and then looking for the corresponding colors among all of the other grayed out files. Again, this coloring helps for categorization of file types and for locating file types (other tasks that we defined). See the following figure which shows the image filetype palette in use. In general, the flat color scheme was disregarded since it was so garish and made navigation much more difficult than when the cushioning was turned on.
Filtering functionality in SequoiaView is somewhat weak insofar as it is not dynamic. One can filter on file size or date which is of course useful when looking for recent files or for files taking up a lot of space. However, dynamic query capability would bring the tool to a whole new level of usability. When someone is browsing for a certain file from a given time period, he may only have a fuzzy idea of what the date is and thus interactive feedback on the date filter would be quite helpful. The implementation of interactive feedback may not be feasible for large file systems however. I found the different types of highlighting and selection interesting. The option which would highlight not only the file, but also its entire path up the hierarchy by using a graduated color scheme was helpful. It lends context to the file and where it is situated in the larger directory structure. I feel that this highlighting mechanism gives more insight into the hierarchical structure than the other highlighting options. Additionally, the tool tips in general provide an
essential part of the visualization since it is through them that you can see exact where in the hierarchy you are.
As far as the interface in SequoiaView is concerned, there are a few things that would improve it immensely. Most notable is that the context of a directory is lost when it is zoomed in. Smooth zooming would help a lot in orienting the user so that he sees where within the hierarchy he is zooming to. Also, I found it cumbersome to have to execute two clicks in order to zoom (right click for context menu and then selection from that menu). As the visualization is specifically for file systems it would also be very nice to see the capability in the interface for actually making changes to the underlying file structure (perhaps through drag and drop functionality). As for insights gained from the tool, I did see that I have a very large ~ 2GB pagefile. I also saw that I had forgotten to delete a very large (~ 1.5GB) directory on my desktop full of uncompressed video frames. I think that the primary strength of this tool is that big files and directories tend to jump right out at you. Thus it s easy to see what s taking up so much space on your drive and perhaps help you manage that better. An additional strength is the ability to locate categories of files (such as image or audio files) through color coding.
filter especially novel since I could imagine a user task involving knowing the program used to create a file, but not remembering where it was saved. All in all, I found Grokker to be a very effective browsing tool. It combines an intuitive zooming interface, detailed tool tips, and even contextualized/ rendered images within the visualization to provide multiple layers of information in the course of browsing. For instance, when searching for Depeche Mode on Amazon I could see the interrelationship of all Depeche mode merchandise such as CDs, DVDs, and books which made it consequently easier to browse. The rendered icons were very helpful in this case since I could recognize certain albums based on cover artwork. Despite the easy interface, I m still not convinced that I should give up my regular web browser though! Perhaps a weakness of the system is that it attempts for form categories in web search information which can be inaccurate. I grokked Diakopoulos and at first was impressed that it had created separate categories for me, my brother, and my father. However, upon closer inspection many of the classifications within these categories were in fact wrong, which is obviously misleading. At the same time, the filtering was again impressive because I could filter based on web domain and see that all the references to Diakopoulos coming from Germany were for me (apparently I m the only Diakopoulos on the internet in Germany!). In summary, I found Grokker to be a much more general, polished, and functional program than SequoiaView, which is logical given that Grokker actually costs money to buy. Though the interface was a little bit rougher in SequoiaView, its specialization on filesystem visualization makes it a useful tool for tasks in that domain. Grokker however, supports more user tasks insofar as it can also be used to visualize amazon.com product information and general web search results. Overall it does a better job at browsing and filtering than SequoiaView.
This document was created with Win2PDF available at http://www.daneprairie.com. The unregistered version of Win2PDF is for evaluation or non-commercial use only.
