This is a summary of the paper I read on Explorer, this is an visualization tool that facilitates sophisticated data exploration with interactive controls to drill-down through hierarchical levels of data. The tool is heavily inspired on the “streamgraph” visualization of Lee Byron’s own listening history. The visualization of Lee Byron had some shortcomings and the authors of this paper thought it did not realize the full potential of’s data. The visualization of Lee Byron had a less-than-satisfactory experience: the graph is very wide and as it is a fixed rendered image, there is no way of seeing more data than what is immediately visible in the image.

The first shortcoming of the visualization was it only shows one level of a users’ data: artists listened to. Music data typically has four levels of hierarchy that users understand:

  • genre
  • artist
  • album
  • track Explorer implementation Explorer attempts to resolve these issues through interaction like being able to filter the stacked graph of a single users’ tag listening data, viewing of different levels of hierarchy, … Explorer allows a limited amount of social context by allowing comparisons between two users’ history in parallel. Explorer also uses animation heavily to make changes in visualization state clear. Humans are much better at perceiving differences in the position of objects as opposed to length, this is applied by Explorer by the use of multiple visualizations as well as interactively allowing a user to re-arrange the stacked graph.

To solve the problem of the limited overview and navigation in Lee Byron’s visualization they added a pair of arrowhead shaped handles on a slider bellow the main stack graph, allowing the user to adjust the left and right limits of the graph by data. To make this time slider less confusing, they draw a small version of the graph above the slider. An example of interactivity is when a user hovers over an object, the object is highlighted and a tooltip is shown. Another way they implemented interactivity is by allowing when clicking on a layer in the stacked graph this layer is switched with the bottommost layer, this allows for any layer to be viewed with a flat baseline for better perception of changes in value. Double- clicks makes the visualization filter based on the object clicked and switch to the next level down of hierarchical data, e.g. clicking on tag rock the visualization will show all artists of the genre rock. The filters applied by double-clicking can still be removed.

Results and discussion

In the results and discussion about the visualization they tell that people familiar with’s data find the visualization exciting and interesting and I must agree on this fact. Another conclusion drawn is that stacked graph display seems to be a popular and accessible visualization for this kind of data but a drawback is the difficulty of comparing two different layers. A line graph doesn’t suffer form this problem but has other issues by allowing respositioning of single layers in the stacked graph they attempt to mitigate these problems. The line graph they implemented ran into several problems which limited its usefulness, it had the following problems:

  • clumping of data in the bottom of the graph, they tried to mitigate this by displaying it in a logarithmic display
  • logarithmic scale made it very complex and hard to view the graph
  • playcounts are integer numbers, it is not uncommon to have more than one node at the same point, this makes the graph difficult to read

The interactive performance was sometimes difficult to maintain, the main cause of this was the way API processed requests and limitations of the Flash platform.

Future work

The paper also suggests some future improvements, including track lengths from the MusicBrainz database is one of those. This would enable the visualization to combine track length data with’s listening data. Allowing visualization of the actual time spent listening to specific track, artists, and tags. At the moment doesn’t use the colors for the same elements in different visualizations, this is confusing and a significant shortcoming. As mentioned before the line graph suffers from a number of shortcomings.

In the future the visualization would like to implement the technique used by  for synchronizing application state with unique arguments appended to the application’s URL. This allows for sharing the link of your visualization in a specific state with others, enabling more social exploration and discussion.


When I first applied this visualization to my own user data of I was amazed. There are so many things you can learn and discover from your own user data. This visualization is really user friendly this is because it offers great ways to interact with the visualization like one click feature to put a layer at the bottom of the visualization. In general this visualization is a great improvement of the visualization made by Lee Byron.

What I learned from this paper is that a good analysis of the data your visualizing and the functionality you want to offer is a must. I like the way this visualization approached the hierarchy of the music data, starting the visualization from the top and by applying filter allowing the user to go down in the hierarchy. Also the interactivity of a visualization can really improve the user experience but also the way you analyze the your own data.

The repositioning of a layer to me doesn’t solve the problem of comparing, maybe this could be solved by allowing users to select 2 or 3 layers and change to another visualization for example a line graph that might allow better comparison.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: