ComicBot: What’s it all about?

There’s been a lot of talk on here recently about our comics-related Capstone project at MLLL.org and what we’re doing with it. We finally decided on the name ComicBot, and presented it at the University of Washington iSchool’s Capstone event a few weeks ago. It’s essentially a comic book recommendation engine, or discovery tool, based on a custom readers’ advisory taxonomy for comics that we built. It’s sitting on top of the MLLL (pronounced “mill”) collection, a student body owned and operated comic book library at Reed College.

The Taxonomy:

We started this project over a year ago in a taxonomy construction class. We all loved comics, and I had been in charge of the comic book library at Reed when I was an undergrad there. Four of us worked together to figure out a set of terms for this taxonomy that would replace subject headings by getting at what a comic is like to read, rather than what it’s about.

Most people don’t just like one kind of thing. I, for example, love both Joss Whedon and Dickens. Even though they might not look like they have a lot in common at first glance, both of those creators tend to have great ensemble casts, and I love that. We tried to get at those commonalities with the taxonomy, to get people between genres within comics and read things that they never otherwise would.

We’ve put it through heavy revisions now, and it has terms describing all aspects of a comic’s reading experience, from the artistic style, to the emotional effect is has on you, to the plausibility of the world in which it takes place, to the common character dynamics at work. Want a beautifully-inked, adrenaline-filled, heart-wrenching, realistic story? The taxonomy will tell you where it is. (Incidentally, it’s Strangers in Paradise, go read it.)

The Process

Through the whole development process, though, we discovered that our users didn’t want a search engine, they wanted a recommendation engine. So what ComicBot does, is after you’ve read and loved all of Strangers in Paradise, it lets you look the comic up by title, and it recommends other comics like it. You don’t have to know what part of it you liked; it’s already been indexed and the taxonomy behind it will get you to something similar.

Technical:

The current instance of ComicBot at MLLL.org, though available to view, is still in alpha, and will be in beta release in August. It’s built on Drupal 7, with heavy reliance on advanced taxonomy modules, the Google Books API, and Similar By Terms, which uses the assigned taxonomy terms to generate the recommendations. There’s a lot of manpower going into the indexing –it all has to be done manually– but the real core of the enterprise is the taxonomy, which is holding up beautifully. It’s a fully responsive design, which should work equally well on a smartphone as on a desktop.

Further applications:

The taxonomy and the indexing work we’ve completed could work in any system to move readers between works and let them find the next great thing to read. As far as we know, it’s the only system to ever look at the visual appearance of the art in comics and extrapolate meaningful groupings across genre lines. For example, in our taxonomy, you might find Charles Vess’ work and Taoko Nakeuchi’s under the same term. Even though they come from wildly different traditions and work with different literary conventions, they share an aesthetic and will often appeal to the same kind of reader.

We want to hear how everybody else is getting readers to their next comic, let us know!