Big Data and Global Trend Monitoring: The Language Gap
Data traces left by humans both on the internet and elsewhere allow marketers to find trending products; organizations can detect potential health hazards and epidemies; researchers can gain valuable insights on the way ideas spread; and so on. However, all of this depends on that they have access to a more or less complete dataset for the places or topics they are interested in.
Problems with acquiring fuller datasets
For one industry ‘trend’ is the operative word. Fashion, of course. Giants like Editd and WGSN are scaling the heights in fashion data analytics, empowering retailers to better control their prices (a practice that may very well translate to ‘less discounts’) and control their inventory. On top of having billions of data points on inventory, its past performance and prices from countless web stores, their Social Monitor feature delivers real-time updates aggregated from more than 800,000 fashion experts and influencers.
The inventory data points give real power to fashion experts across the globe. As for the Social Monitor, though, one might ask: how many languages does it speak? That is, it might be prudent to guess that their social trend insights are bound steadfastly to the English-tweeting world. If this doesn’t alarm you (it probably shouldn’t), implications of a graver nature following:
In an intriguing article on Foreign Policy, Kalev Leetaru states how HealthMap, the global, news-powered powered disease mapping project, actually missed the Ebola outbreak by a day after the World Health Organization was notified. The problem? “Ils ne parlent pas le français.” They [Big Data] don’t speak French. In many countries one or two local newspapers are published on the internet in English, and only minuscule parts of the rest are translated. Most never are.
This is a serious hindrance for global Big Data initiatives of any kind. You can yield useful translations with Google Translate, but you can also get rubbish. A workaround could perhaps be found (at least in the Ebola outbreak case) in developing computer-readable standards for publishing important news concerning health hazards; but who’s to say that anyone would be willing to comply? Another workaround lies in crowdsourced translation, like in the Duolingo app where collective understanding provides accurate translations.
A simple solution doesn’t exist, but, as more and more nations join global data initiatives and develop their own, the language gap will be closer to being bridged. Some might object that the real beauty lies in that it never will.
By Lauris Veips
