A Note on Usable versus Unusable Crowdsourced Tags

I’m going to be posting anywhere from 1 to 3 articles using animenewsnetwork’s tag system over the next week or two, so I think now would be as good a point as any to make an important point about when it is and isn’t ok to use.

There’s lots of interesting work to be done on the performance of anime and manga of different genres. Classifying the genre of a work of entertainment is tough. Especially when the definition is as amorphous as it is for something like moe. However, in practice, it actually isn’t a huge deal if a site is consistently applying the same incorrect definition; even blatantly incorrect definitions (I’m looking at you, shonen=the genre containing battle action) can yield meaningful results about the type of series fitting the definition being used if they’re consistent. Dealing with them just requires a willingness to dive in and determine exactly what your numbers say.

A much more common and hazardous pitfall tags slip into is one I call selective incompleteness. Quite often, via either lack of attention to the tag or too much attention paid to too few cases, whether or not a series falls under it will depend on whether it’s popular or not. This is a problem for lots of reasons when trying to break down the sales of a genre; every Genre A series that aired over a 10-year period might come up with worse average sales than the two most popular Genre B series of that period, even if the Genre A was generally much more profitable.*

The above is an extreme example, but I hope you see my point. If we’re going to talk about the average performance of a genre, we want their averages (or totals), not just a selection of more-popular-than-average shows that happen to be from those genres. Since popularity can correlate with sales, the risk of higher-selling shows getting selectively tagged is real, and worse so the smaller the tag is.**

ANN’s moe tag, while not quite that extreme a case, is a victim of selective incompleteness to a dangerous degree. One of the easiest ways to test the veracity of a tag is to check whether or not it’s applied consistently in-between seasons of a show. If a series gets a second season largely in line with the first, they should be tagged equivalently. If there is selective tagging going on, then sequels are probably not the only victims, but they are the most visible and arguably objective sign that it’s happening. But, Hidamari Sketch and Nogizaka Haruka no Himitsu both fall victim to selective tagging under this tag. Specifically, only the first season of Hidamari Sketch is classified as moe, while only the second season of Nogizaka Haruka is. Neither of these series had sequels that radically changed gears.

I could point out other things that are suspect about the tag (like how Kyoto Animation produced something like 30% of all moe between 2005-2008), but the sequel thing is a red flag. In general, I’ll give tags some leeway, given the fact that some classifications are difficult to make. But any label that splits the difference between seasons of the same show (excluding reboots like Im@s or spinoffs like The Unlimited) is obviously too inconsistent to use. As a rule of thumb, any tag which I can point to as containing several cases of franchises being tagged inconsistently is one that I will not be using for any meaningful analysis.

*Since the profits of the anime hinge on the top 10% of series, the precise designation of a blockbuster like Bakemonogatari do have a lot of analysis-swinging potential. You can’t just ignore the exceptional cases (producing those franchise-series outliers is the business of most publishing industries), but their swing potential is the biggest reason why accurate classification is important. The fact that some of these series have disputed status that varies between databases is the lynchpin of an upcoming addendum to one of my older pieces of work.

**”Smaller” tags also carry a secondary issue, the potential of carrying a definition so specific that it’s hard to draw general conclusions due to small sample sizes. Any list less than 20 shows in length is tough to be general about.

Advertisements

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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