\

Feedback from mid-term critique

 In In Progress - Thesis

The feedback from the crit was just absolutely incredible. There were a lot ideas brought up that can just push this project even further.

For the web version

  • Any way to include song samples?
  • The graph is overwhelming at first. The labels have a lot to do with that, especially since only half of them are displayed while the other half are not. It should reach a point when a user zooms in, they should all show up at the same time.
  • A toggle to turn the labels on and off could help substantially
  • The white type is slightly jarring. I agree with this. I should definitely tone it down a bit.
  • The information bar on the left should give a description as to what the size of the nodes indicates.
  • Having a home page that explains this graph could be extremely beneficial. I’ve already started working on a site that will house this visualization. My plan, as mentioned in a previous post, is to do more visualizations that just this one. I want them to be all focused around music.
  • Writing user case studies can help differentiate what users can gain from this. A well educated person in music may find this information fascinating, but may have difficulty using this (as I’ve already experienced). ¬†Someone familiar with network analysis will have no problem with this.
  • Would it be possible for a user to select more than one node? (SUCH a good idea). In the config.json file, I’ve enabled the groupSelectorAttribue which allows every node of a single color to be grouped into their own groups. Since all the songs share one color, they are formed into a group. When clicking it, only the songs display on the screen. I feel as though I can push this further by allowing more than one group being able to be displayed. I have also enabled the search bar which allows information to be searched. However, is this beneficial if one doesn’t know what to search for?

For the print version

  • The version that displays the connections as lines is not really understood at first but has potential. I had another one up that was exact copy of the web version but i feel that isn’t as interesting.
  • However, the impact of what the connections as line is displaying is better understood. It’s better to view it at once this way. Perhaps going even bigger can help this (side not: I spoke to someone from photo about using those large printers. I can get a 100 ft roll of paper for about $50).
  • The colors is another issue. 18 colors the limit that I should not go beyond. To help ease the amount of color, I tried using shades of each color. The green, while it looked good on screen, printed like crap. Either way, they were still too close in color.
  • One of the biggest questions with this is what is the best way to display specific sets of data? While pie charts and bar graphs may help, it’s not really interesting. Another graph I had printed showed specific categories as connections. The ones I had printed out were the BPMs and the country of origins. With this, I can compare sets of of attributes to each other. A series of posters could come into play with this. Male and female can be displayed on one poster. The BPMs can have another. Why is this better than the pie chart? Because you can see the context of how many of the songs are actually connected because of the one attribute.

As I said, great feedback. While I was writing this, I was trying to figure out how to change the text for group labels. The default is set to Group 1, Group 2… etc. So instead of the songs being labeled group 2, they are now labeled as songs. The fix wasn’t too bad (someone else on the interwebs had the same problem). This line of code had to be added to the main.js file:var groupNames = ['Group One','Group Two','Group Three','Group Four','Group Five','Group Six','Group Seven','Group Eight','Group Nine','Group Ten'];

which I now have changed to this: var groupNames = ['Decades','Country of Origin','Songs','Genre','Gender','Key Signature','Length of song','Whether frontman plays instrument','Needs to be fixed','Beats per minute','Group or solo artist','If the single won a Grammy'];.

This is how to set labels for the group selector. You still have to tell the javascript to actually use these labels, so this: var a = [],
b,x=0;
for (b in sigInst.clusters) a.push('


‘ + groupNames[x++] + ‘ (‘ + sigInst.clusters[b].length + ‘ members)

‘);
$GP.cluster.content(a.join(“”));
gets added in replace of the old one. What does it mean? I have no idea. How does it connect to the custom labels I set? Looks like it’s using the groupNames tag. I’m happy this worked though. Gives me more to work off of. My next steps are to continue working on the print and to get force atlas working. I’ve located the version that works with this version of sigma I’m just having problems actually getting it to work.

The other thing I’d like to do is to space the nodes farther apart. No matter how far I push them away from each other in Gephi, it still exports as the same layout. I’ll have to dig around a bit. Maybe once I get force atlas to run, it will help with that substantially.