|
|
| Line 1: |
Line 1: |
| '''<font size="6"> SparkClouds: Visualizing Trends in Tag Clouds </font size="6">'''
| | http://www.clomidlmt.com/ clomid online - order clomid no rx |
| | |
| I found a very cool IPAD app that would seem to have some good application for PT sedntuts (and others as well.) This allows annotating on PDF files- there is a highlighting tool, a marker, a text box option, and an eraser. The app is called noterize and it costs $2.99. So far I have been able to annotate in the IPAD and then put it in drop box so that I can access from my tablet. Excellent!
| |
| | |
| =How to design SparkClouds?=
| |
| | |
| ==Basic reasons for SparkCloud design==
| |
| Bongshin Lee et al. based their SparkClouds design decision on two scenarios:
| |
| * Jason, a stock market analyst, is dealing with a lot of tag clouds to figure out what for example the latest trend in mobile communication is or what’s gossiped over right now. Jason needs a high level overview of trends.
| |
| * Laura is a researcher, working in a team of 16 people, who wants to track her team’s activities changing from week to week. She needs a time point by time point view of tag clouds to gain information about what was done at a specific date by her team.
| |
| | |
| Due to these two needs SparkClouds were designed by combining standard tag clouds (showing tag popularity by font size) and sparklines, which are simplified line charts (showing the trend over time).
| |
| | |
| ==Tag Clouds as used in SparkClouds==
| |
| Here two design parameters are given -> Font Size Encoding and Layout
| |
| | |
| * Font Size Encoding: To support both stated above scenarios two methods to calculate the font size encoding were defined. The first one encodes tag font size by the frequency of its appearance over the entire period. This supports Jason’s need to see the most popular topics over the whole time span having the largest font size. The second encoding method supports to view the most popular topic, in Laura’s case an activity of her team, at a given time point. For this method the font size for each tag is stable -> it does not change whether one or several tag clouds are shown in the visualization. This allows comparison between tag clouds only in rank but not frequency. SparkClouds, by default, use per time point font size encoding.
| |
| | |
| * Layout: In contrast to conventional tag clouds, SparkClouds sort tags in alphabetical order, which makes them still exchangeable with any existing tag cloud because they still have the same basic structure.
| |
| | |
| ==Sparklines used in SparkClouds==
| |
| For being one of the most common visualizations to show trends over time, line graphs were chosen to support SparkClouds need of time point by time point interpretation. These line graphs were simplified by eliminating both axes and changing them into sparklines. Still the “vertical axis” shows the relative popularity of the tag and the "horizontal axis" the time span. To easier spot periods in which a tag was not used the area between the trend line and the hidden horizontal axis was filled with a light gradient blue color. This feature will help to figure out the stability of a tag or its climax. Sparklines were also chosen as visualization type because the can easily fit adjacent to tags without cluttering the layout.
| |
| | |
| ==How to unify Sparklines and Tags?==
| |
| [[Image:allignment.jpg | 300px | thumb | '''Figure 2''' : '' Design Varieties to unify tag clouds and sparklines'']]It was needed to form sparklines and tags into one single unit of information to make the unit easier to recognize for the user. Following possibilities were worked out:
| |
| | |
| * Alignment: A Sparkline is placed before, after, above or below the tag. Further a handwriting like font was used and alignment of both text and sparkline on the baseline was tested to provide visual continuity but the legibility of the tag was compromised.
| |
| * Overlay did not work because the tag seemed less readable. Also problems occurred due to the varying word length, making it difficult to compare points over time on sparclines of different scale.
| |
| * Mirrored sparklines as shown in Figure 2 were often interpreted as non mirrored ones causing an opposite understanding of the trend.
| |
| * Foreground and background colors: It was tried to color one tag and its sparkline in one color and the following unit in another, further applying this pattern over the whole SparkCloud. It still looked clutter and was again out ruled because of the fought to use color in a more effective way.
| |
| | |
| Finally it was decided to fit a circle with a faint gradient blue (from a colored center to white periphery) in the background of the tag and sparkline to form one single unit of information. To ensure readability tags were outlined by a thin white border.
| |
| | |
| =The experiment=
| |
| The goal was to figure out if SparkClouds would outperform three alternative visualizations in efficiency of usage.
| |
| | |
| *Four different Visualizations
| |
| | |
| The compared visualizations were Stacked Bar Charts (Figure 3), Multiline Graphs (Figure 4), Parallel Tag Clouds (Figure 5)and SparkClouds (Figure 1). Throughout all Diagrams words were presented in alphabetical order and the same range of font sizes was tried to be implemented.
| |
| | |
| {| class="wikitable"
| |
| |- valign="bottom"
| |
| | [[Image:stackedbar.jpg | 200px | thumb | '''Figure 3''' : '' StackedBar '']]
| |
| | [[Image:multiline.jpg | 200px | thumb | '''Figure 4''' : '' MultiLine '']]
| |
| | [[Image:ParallelCloud.jpg| 200px | thumb | '''Figure 5''' : '' ParallelCloud '']]
| |
| |}
| |
| | |
| | |
| *Tasks and Date
| |
| For the 75 most frequent words from the first chapter of 5 well known books, the number of occurrence was randomly generated. This was done for each word for 12 months. The top 25 for the selected month were highlighted.
| |
| | |
| The four visualisations were tested across six tasks.
| |
| {| class="wikitable" border="1"
| |
| |-
| |
| ! #
| |
| ! Task
| |
| |-
| |
| | 1
| |
| | What topic was ranked as the most frequent topic (top-ranked) during month X?
| |
| |-
| |
| | 2
| |
| | During which month did topic "A" occur most frequently?
| |
| |-
| |
| | 3
| |
| | Which of the following topics experienced the longest continuous increase/decrease in rank during the year (e.g., for the most months in a row)?
| |
| |-
| |
| | 4
| |
| | Which of the following topics experienced the largest increase/decrease in rank from month Y to month Z?
| |
| |-
| |
| | 5
| |
| | Which of the following topics was the most frequent topic (top-ranked) for the greatest number of months during the year?
| |
| |-
| |
| | 6
| |
| | Which of the following topics was ranked among the top 25 topics most consistently over the year (e.g., was ranked at all during the most months)?
| |
| |}
| |
| | |
| *Study Design
| |
| Measured were task completion time, accuracy to estimate the efficacy of each visualization and participants’ subjective preferences for each visualization. Participants were further urged to answer questions as quickly as possible without sacrificing accuracy. At the end of the experiment, participants were asked to rank the visualizations in order of preference, from most to least preferred. The experiment lasted approximately two hours
| |
| | |
| *Result
| |
| As shown in Figure 7 SparkCloud and MultiLine Diagrams have faster overall task times.
| |
| | |
| | |
| [[Image:overall task time.jpg | 200px | thumb | none | '''Figure 7''' : '' Overall task time'']]
| |
| | |
| | |
| The performance differences among the four visualizations within each tasks is show in the next diagram (Figure 8). White dots in columns indicate the total number of significantly slower visualizations within the same task.
| |
| | |
| | |
| [[Image:task time by visualization for each task.jpg | 200px | thumb |none | '''Figure 8''' : '' Task time by visualization for each task'']]
| |
| | |
| | |
| The diagram below (Figure 9) states participants personally preferred diagram. Although it still received a low rating regarding the overall appearance. SparkClouds still seem to be looks cluttered.
| |
| | |
| | |
| [[Image:participants ranking.jpg| 200px | thumb | none | '''Figure 9''' : '' Participants ranking'']]
| |
| | |
| =Conclusion=
| |
| In general it is to say that SparkClouds were the most consistency competitive visualisation. It was only once outperformed by a Multi Line Diagram. SparkClouds inherit both sparkline and tag clouds advantages; Sparklines showing the trend on very little space and the compact and pleasing layout of tag clouds. The easy way to replace current tag cloud with new SparkClouds was also mentioned as an advantage.
| |
| Negative mentioned was the lack of details of the sparklines. Participants estimated their values. Also SparkClouds inherit the weakness of tag clouds that longer words receive more user attention because of the occupied space.
| |
| | |
| =References=
| |
| *[Bongshin Lee et al., 2010] Bongshin Lee, Nathalie Henry Riche, Amy K. Karlson and Sheelagh Carpendale. [http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=5613457 Visualizing Trends in Tag Clouds]. IEEE Transactions on Visualization and Computer Graphics (TVCG), vol. 16 no. 6: 1182-1189, 2010, November/December 2010.
| |
| | |
| | |
| [[Category:SparkClouds]]
| |