My motivating question is how did http://wikimapia.org grow over time? Like wikipedia, wikimapia is another social experiment to see how well people can collaborate with each other, except now with the locations and names of places. This data is useful for answering questions such as 1) what countries contributed to the initial growth of wikimapia the most and 2) how did certain key terms (such as schools and colleges) grow on the map over time? Analyzing such trends may lead to better insight on the nature of user interactions.
There were some interesting findings from the visualization:
In general, the data grew starting from Moscow then it quickly spread to Europe (especially Great Britain) and to the United States (mostly East and West Coast). These are English-speaking, wired areas so it makes sense. Strikingly, the site is not very popular in Africa, South America, Asia, Canada, Middle Eastern countries. These areas have almost no data plots.
We can also analyze how certain places, restricted by certain filter keywords, grew over time. As an example,
the keyword "school" has interesting growth. It again, starts out mostly in Moscow. It grows steadily in the Eastern part of the United States. Many of the Asian countries such as Japan seem to put their school first on the map. It might easily be a symbol of national pride.
Interestingly, the term "street" is mostly restricted to the United States and Great Britain. I guess other countries do not enjoy labeling their streets.
The applet is written in Processing with the data scraped from crawling through http://wikimapia.org/# where # is the id number. I acquired it by running a PHP script that scrapes the page and inputs the data into the MySQL database on my computer.
Simply, when you open the applicaiton, there will be a live animation on the map of all the data points popping up and disappearing. The map will display only a certain amount of points over time as the points will start to fade once they have been put on the map. The points have a certain opacity to them so that if points overlap, they will create a stronger black. Also, if there are many points in the same area, then it will seem to take longer for black to fade out. These perceptual changes will make it easy for users to distinguish patterns on the map. This is an engaging way to view changes over time and it is similar to the Super Bowl Twitter visualization in class.
There will be a filter input box on the top that you can enter key terms in. You can just start typing and it will go into the filter input box. Press Enter to input the search term., then only the points with names associated with that term will be highlighted in red in the animation. If you hover your mouse over the red points, then you will be able to see the labels for those points. Key terms are easy to come up with because of abundance of data. Typing "beach" into the application results in over 10 hits. Try terms such as "james", "church", "street", etc.
You can see how much of the data has been put on the map by looking at the Progress bar on the top. You can also pause the map by clicking "Pause / Resume" and see all the data that is present on the map at that time. You can also click on different parts of the progress bar to shift the map to that state. Also, you can reset the map by clicking on the "Reset Map" button in the top right corner. But the animation will automatically reset once it reaches 100%.
Work to be done:
When you have a filter term, some places show up as red. When you hover these red points, labels pop up. Currently, I crudely hash the ID of the data point to some location around the red point. But inevitably, some labels overlap. There must be a better way to deal with this problem. One method is using a scrollable menu next to the point that you can use to peruse the data.
- There is a progress bar with clikable feature but it would be nice if one could drag the progress bar around to different states, and see the map change dynamically.
- When the filter terms are highlighted in red, the black ones could fade out more to emphasize the red. However, I think red stands out so much already because of its difference in color channels that I don't think it is necessary.
- When the map is paused, the integrator keeps on going so when the map is resumed, it looks as if the map is empty because points have become invisible. There must be a way to pause the integrator state for each point.
The data set is the first 100000 data points on Wikimapia.