As if it wasn’t enough for Macedonia and Macedonians to be erased or added to through forced – and illegal – name changes and more, now we have AI in the mix doing the erasing or adding. This next was new to me, but it’s important to raise awareness of this so that others can join the fight.
When using the auto-translate on the LinkedIn platform, I recently noticed something odd, perhaps nefarious, going on. (I reached out to a friend, a professional translator, who told me that LinkedIn uses its own proprietary auto-translate software; I address Google’s auto-translate, later in this column.) Three examples follow and I have not included the names of the authors of the posts, for a variety of reasons.
The original post in Macedonian followed by the English translation using the LinkedIn auto-translate function:
Example #1
“Далеку од дома, успешни, реализирани и влијателни Македонки во светот. Секоја од нив е инспирација за следната генерација жени.”
“Far from home, successful, accomplished and influential women in the world. Each of them is an inspiration for the next generation of women.”
Example #2
“Предложената Резолуција во Сенатот е признание за Македонците во Америка и нивната улога во унапредување на САД во последните 150 години”
“The proposed resolution in the Senate is a recognition of Americans in America and their role in advancing the United States over the last 150 years.”
In these first two examples, “Macedonians” have been left out; in the first example, Macedonian women – not just “women.” In the second example, “Macedonians in America” becomes, weirdly, “Americans in America.”
Here’s the third example in which the AI algorithm inserts its own political bias:
“ДАЛИ ЗНАЕТЕ ... Во Македонија во почетокот на 1950-те години во Скопје била формирана фабрика за воздухоплови во која се изработувале едрилици кои биле меѓу најдобрите во светот.”
“DO YOU KNOW... In North Macedonia in the early 1950s, an aircraft factory was established in Skopje to produce gliders that were among the best in the world.”
The author clearly wrote “In Macedonia…” and yet the auto-translate turns it into “In North Macedonia….” In this case, it’s not erasing Macedonia or Macedonians, but adding to them something they reject, the adjective “North” making them, in many cases documented here, “North Macedonians,” a people that do not and never have existed.
What the bloody hell is going on here?
Next, let me turn to the Google auto-translate function.
I took that second example, from above, and plugged it into the Google auto-translate function and discovered something entirely new to me. Google translate offers a “tone” right below the box that displays the translated words. It offers three settings, Standard, Casual, and Formal.
When plugging in this paragraph….
“Предложената Резолуција во Сенатот е признание за Македонците во Америка и нивната улога во унапредување на САД во последните 150 години”
…..and using the “Standard” setting, we get:
“The proposed Senate Resolution is a recognition of Americans and their role in advancing the United States over the last 150 years.”
Somehow, the Macedonians in America were “disappeared” to use a euphemism.
But using the “Casual” and “Formal” settings both rendered:
“The proposed resolution in the Senate is recognition for the Macedonians in America and their role in advancing the USA in the last 150 years.”
Weird, huh?
What is the bottom line here? Again, I’m raising awareness and my translator friend suggested that, when you see this going on, you reach out to (in this case) LinkedIn and Google to continually offer feedback until a more robust answer can be found.
For Google Translate, there’s a “send feedback” tab just under the bottom right corner of the translation table, so sending examples of translation errors is best that way for that.
For LinkedIn, it’s reporting the incorrect translation (providing screenshots as proof) at https://www.linkedin.com/help/linkedin/ask/itn
All of this calls for continued vigilance by Macedonians and their friends.
This makes me feel so 😡.