Google Translations Risk Spreading Bad Info, Violates Copyright

I was just reminded about a copyright transgression made by Google against a website I used to publish.

In her On Tech column in the New York Times, “What’s the deal with Google now?”, Shira Ovide discusses the legal cloud growing over Google, resulting from its monopolistic tendencies. Among the several infractions that she cites, Google has for some time scraped information from websites and published it on its search results page, often delivering what the user is looking for and saving that user time and another twitch of her clicker muscle. Good for the user, bad for that other website that, like most websites, counts visitors as an important metric for its revenue model. And, if the information that is scraped is copyrighted, the website receives no compensation.

This reminded me of another service that Google has provided for the last 10–12 years: translation of web pages from English into one of many available languages. I first discovered Google Translate around 2009 or 2010 when I was executive producer of a free-to-the-public, not-for-profit patient education website that explains medical lab tests. Amazingly, all I needed to do was plug a URL into a text field and this Google service would translate our English language page into Spanish or French or German or Polish or just about any language you could think of.

Pretty great, right?

Not so much. There were, and still are, two fundamental problems with that service.

First, the material that I had asked Google to translate was copyrighted. And in fact, we earned revenue by licensing the content of the site to publishers in 14 other countries who translated the content into their national language. Google would never pay us a penny for its translations. Heck, we wouldn’t even know if and when the pages were being translated.

Second, and this has broader implications, Google’s translations were hardly nuanced nor adept at translating medical terminology and concepts in an accurate and compassionate way. Our original content benefited from a rigorous editorial process that captured the practical experience of 20 MDs, PhDs, and Masters-educated practitioners who were also extraordinarily sensitive to how patients and their loved ones received sometimes frightening medical information. As editors, our editorial board of experts were unparalleled in their ability to choose language that helped to minimize the fear and anxiety among our site’s visitors.

Google’s machine translation was horribly limited in this capacity.

Similarly, in translating our pages, our international partners were not just putting the text into a different language, but were adapting the content to differences in healthcare policy, practice, and culture in their respective countries. So someone in Spain who used Google translate to convert one of our English language pages would see something quite different from the authorized Spanish version that was published there.

I tried to alert Google to my concerns at the time, but never had a dialogue with them, and I didn’t have the resources to pursue the matter.

Ovide’s column reminded me that these problems still persist, though I suspect the quality of the machine translations has certainly improved. Nevertheless, the translation service should be part of whatever investigations are occuring into Google’s Internet dominance.