Strategic analysis
Google Translate is Changing Sources

I came across something quite odd today. Earlier I shared a quote from an article from a Danish media site about the degrading relevance of news.

The number of news articles has more than doubled in 10 years, while the proportion of the population who cannot identify a single relevant news within the past 24 hours has increased from approximately one quarter to almost half.

I then, as I always do, also wanted to link to the story, but since it was a Danish story I wanted to share a translated version. No problem, you think, just point people to Google Translate, right?

But then this happened:

The Original article:

Roughly translated it says:

"I think we will look back on 2011 as the year when journalism triumphed itself to death," he told Jyllands-Posten, adding:

Jyllands-Posten being one of the largest national newspapers in Denmark.

But here is the translated version:

Google Translate changed the source from "Jyllands-Posten" to "CNN".

Now, I'm sure this is not intentional. I don't think Google is favoring CNN to the point where they would credit them for a story they have no part in. But it illustrates a potential problem with automatic services.

Google Translate doesn't actually translate each individual words. It looks at word patterns, matching those with similar word patterns in another language. What has probably happened here is that the common phrase referencing Jyllands-Posten in Danish, is, in that sentence pattern, a similar common phrase referencing CNN in English, hence the change.

It is one thing to translate the words, but when you change the places, things or the companies it can lead to all kinds of problems. Imagine Google Translate changing Obama to Sarkozy when translating a document about an important international issue.

The real lesson, of course, is always to check the original (which is often easier said than done).