Rick Falkvinge, better known as the first leader of the Swedish Pirate Party, made an announcement yesterday of an upcoming launch to his news service that would focus on delivering short-form news. So called Heartbeat Stories would be delivered in images over social networks, with ads embedded directly into the image, thus bypassing ad-blocking software altogether. Why I want to talk about it is because the format reminded me of how I used to write anime news back in the day. This new direction in journalism, if it takes off, can possibly lead to automating the process of writing the news. There’s obviously a trade-off in quality, but one news organizations might be willing to take.
The Heartbeat Story format dictates three sentences: Fact. Background. Satire. Obviously the satire part is an editorial choice, other than that short news formats aren’t exactly a new invention. Times are changing and so do news consumers. Microsoft came out with a study that basically said most people today have an attention span comparable to a goldfish’s. This evolution of the human brain, if you want to call it that, was almost necessary in our information society. We’re talking just news here, but there’s so much trash to filter through it’s not even funny. I have a project pending to make a filter for Anime News Network’s RSS feed that would weed out most news that are not about anime. It’s ridiculous that a site supposedly specializing in anime can’t afford to put up an RSS feed with only anime news. We have to wade through all the gaming and general geekdom wankery that they publish before we get to the anime bits. Can’t exactly blame them. Well, I can but, your average anime fan likes E3 announcements and personality obituaries more than anime news.
To give credit where credit is due, fandom has found their own ways of filtering through all the crap. Some have turned their eyes to Reddit, where the community helps out in deciding what’s relevant and what’s not. Another consequence of being on Reddit is that you don’t even click the links anymore, rather you just read a posted tl;dr in the comments section. Or a funny meme as the most upvoted comment, because funny memes are more relevant than news. Another fascinating aspect of Reddit is that people are more likely to check out links with a thumbnail beside it. Indeed, Falkvinge’s format is hitting in the right direction.
I realized years ago that as a data geek I’m not interested in all the ham that’s added to your typical news article. When it came to anime news, I was only interested in the fact part of the story. So I started writing this tiny news blog where I posted news with only the bare essentials. Posts were usually a sentence or two in length, with a link to the source. Writing them took way more effort than I had originally thought, but it was an effective way to compete with ANN in terms of speed. Suffice to say I learned quite a bit about publishing anime news from that endeavor.
One of the more reassuring things about Japanese-to-English journalism is that Japanese publishers have this lovely uniformity to their articles. Not only do their articles have very formulaic structures depending on the publication and the type of news it is, they envelop all franchise titles in brackets, which enables even people with Google Translate-level of language proficiency to translate their articles.
Another fact to consider is that there’s only so many news to publish about upcoming anime titles, which is why we can start talking about automation. The cycle of an upcoming anime title’s life in our news feeds starts with an announcement to an important announcement regarding an established franchise. It continues with the announcement of the adaptation, possibly along with some staff information. Eventually the seiyuu cast gets announced and a key visual posted. In the final stages they put out a promotional video and home video release information. That’s it. We’re left with five to seven news classes that we can teach a computer algorithm to discriminate between and make a prediction model for the task. Since there are only so many different text patterns we can employ, we can set up post templates for each news class. When the news hits, our algorithm decides what type of news we’re dealing with, then a script finds the franchise’s title, the relevant dates or other class-relevant information and inputs them into the appropriate template. Finally, the article is published with a confidence rating that our algorithm spits out. Users or service maintainers can then put as much trust into the article based on the rating, and correct manually if necessary.
Obviously, the real world of news publishing comes with a ton of exceptions, but I’m confident anime news publishing is simplistic enough to be able to solve or bypass them to a high degree. Besides, most fans don’t give a shit what the news say, they’re only interested in the final product as it comes out. The primary function of anime news is to create hype. But if that’s all anime news is, it makes very little sense to waste time writing it. In a few years’ time I can see ANN semi-automating most of their news articles. I can also imagine reading them from a fucking JPEG or GIF.