The Social Structure of the Anime Industry

How does the anime industry look like if we try putting its abstract social structure into form, without knowing a whole lot about it? I posted a web app this morning that tries to accomplish just that. It’s a bit simpler than the image render below, but it’s got the benefit of interaction. Be patient, loading may take a while!

Anime Creators Network
The Social Network

So to make a long story short, I tried visualizing how staffers from different anime projects are connected to each other. I wanted to see if there is structure to their social network and how that might benefit our understanding of certain communities forming within it. The process involved querying ANN’s Anime Encyclopedia for relevant staff information. All that was needed after that was a bit of grunt work to build the social graph. Each dot in it represents a professional who is credited as having worked on anime. Each line connecting any two dots is to imply that the two professionals represented by them have worked on at least one anime together. In graph theory, the dots and the lines are called vertices and edges. In this graph visualization, dot size increases proportionally to the number of edges that a vertex has. Color intensity of edges tries to convey the level of collaboration between two persons, meaning that if they worked on many shows together, the color of their edge would be more intense.

While it would be nice to know just how much interaction there is between staffers of different companies, I’ve taken the liberty to simplify the model by assuming that everybody who has been credited for a certain anime title knows everybody else credited for that title. In the real world, an in-betweener working for a subcontractor is very unlikely to ever interact with an anime’s main producer, for example. But I don’t think it’s too hard to imagine that it would not be impossible for him to do so if he wanted to. It’s funny to think about these loosely defined connections, because in my country of Slovenia most favors and business transactions are facilitated by knowing a guy who sort of knows the guy who has the thing that we want. In light of that, this graph should be taken as a low-effort approach to solving the problem. For serious researchers I suggest strengthening the dataset with more and better data. The edges especially need more relevant attributes.

Some of the questions may be a bit too hard to answer without looking at the raw data though, so let’s do that right now. Who are the most connected people in the industry anyway, going by the metric that we just described? Who can make favors happen the fastest?

  1. Tsuruoka Yota (sound director): 1448 edges
  2. Sakata Junichi (director): 1426 edges
  3. Mima Masafumi (sound director): 1399 edges
  4. Taniguchi Moriyasu (animation director): 1360 edges
  5. Okazaki Yukio (episode director): 1332 edges
  6. Honda Yasunori (sound director): 1328 edges

To me it’s not all that surprising that the most connected people would be directors, but them being sound directors I thought was interesting nonetheless. Still, there are other connectivity metrics to consider. If we wanted to know how easy it would be for a person to contact anyone from the industry, anyone at all, regardless of the degree of separation, we would need to think of something else. Here, a closeness centrality measure comes in handy. Taking that over node degree measure, the ranking above doesn’t change all that much, but a producer named Okuda Seiji ladders up and passes Taniguchi.

The web app requires some patience to work with. It’s a bit slow due to the fact that the browser needs to draw whopping 11828 vertices and 563177 edges from a 50 MB dataset on a single thread. Do note that the graph omits voice actors and foreign language release staffers. The former mostly due to the fact that my computer’s memory is limited, not to mention all desktop software that I tried had problems visualizing more than a million edges. The latter because I wanted to focus on the Japanese side.

If you’re interested in the code that I’ve written for this project, take a look at the GitHub repo. I used Gephi for visualizing the graph. As always, comments and errata are appreciated.


Leave a comment

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