Moderation In Foreign Languages Is MUCH Harder Than You Think

I have most likely learned and lost more language than you have. I’d listen to my opinion if I were you.

Moderation is hard. The moderation team on any platform is subjected to a fire hose of crap aimed straight at their face every day. They see stuff from people who think nothing of posting stuff like If you don’t know what this is, consider yourself lucky. This is still not the worst, however. People will post death threats, hate speech, child porn, etc. When you moderate, you have no choice but to be subjected to this crap. As an end user, you can block. As a moderator, you cannot.

Now let’s talk about myself. The only two languages that I can claim to speak fluently right now are French, which is my mother tongue, and English. When I say French is my mother tongue, I’m taking a shortcut. My mother tongue is actually a dialect of French-Canadian French. However, we do learn an international form of French in school. Therefore, saying that I speak French is not a lie.

I’ve learned and lost a ton of other languages over the years. You want a list? Here it is, in no particular order:

  1. Spanish
  2. Portuguese
  3. Italian
  4. Norwegian
  5. Swedish
  6. Latin
  7. Attic Greek
  8. Hindi
  9. Sanskrit (I also taught it.)
  10. Classical Chinese
  11. Mandarin
  12. Classical Tibetan
  13. A special form of Japanese used to annotate ancient Buddhist texts.

I’m probably forgetting some languages that should be in this list. The “and lost” part is important, because I’ve lost some or most of my ability in all these languages. Part of it is due to disuse. Part of it is the cancer that was in my brain. Chemo brain is a jerk. I’d say Spanish and Portuguese are not faring too bad. However, my ability at Classical Tibetan or that special Japanese is probably laughable right now.

“Oh, but you can just…”

No, you cannot. Not if you want to do a decent job at moderating your platform.

“Oh, but you can just shove the text into a translation tool.”

Even if the translation tool is integrated with your instance. This is additional moderation time. The number of moderators is not infinite. Consequently, you want to make it easy for them to do their job. However, there’s a bigger issue here. Look at this crap:

This is a screenshot of Google Translate doing a shit job. Read the rest of the article for the description.

What I gave to Google is something I would be likely to say. Google translated toé mon osti de chien with you my dog osti. Google does realize that the non-standard spelling toé is you. Note that I’ve tried other variations, and sometimes it does not know what to do with toé. However, it does not know what to do with osti. I’m really wondering what a moderator would think when faced with this translation. What does osti mean? It is a verb that means to like and then the entire sentence might mean you like my dog or something similar?

Not at all. This is an insult and is immediately recognizable as such by the native speakers of the language. This is exactly what would come out of my mouth if I was angry at someone, and I was speaking in my dialect. There is no direct translation that I can think of that makes sense in English. I’d translate it as you fucker.

Yes, a moderator willing to spend the time could figure out what to do with osti. However, we’re trying to keep the moderation burden low, and this extra work is not furthering this goal.

Okay, let’s try something internationally recognizable:

This is a screenshot of Google Translate doing a shit job. Read the rest of the article for the description.

Now Google is translating Quel con! with That with! but that’s because it misidentified the language. It thinks we’re speaking Italian. If we force it to understand French, it will give the answer What a jerk! which at least is in the ballpark. I would have translated it as What an idiot!

The examples I gave above are extremely simple, and they barely scratch the surface of the problem. We’re not dealing with word play, backhanded compliments, etc. When you have to bring into play the culture of the foreign language you have to deal with, then the problem becomes very complex very quickly. The translations tools are not as useful as you think they are.

“Okay, but you can just hire moderators who speak the language.”

As if… Since the notion is to hire native speakers of the language, well, okay, that bit of the problem is handled perfectly. However, what makes you think that those native speakers are going to be able to moderate well? I don’t want to put into a position of power people for whom I don’t have some evidence that they are going to be good candidates. The fact that they were a moderator elsewhere does not count for anything. They could have been moderator for a platform that promoted hate speech. Yes, these shitty platforms have moderators too, believe it or not.

Okay, so let’s suppose you have one person on staff who is able to deal with French-Canadian dialects. What happens when that person temporarily unable to do moderation. Do you tell people “Hey, until Pierre is back, you cannot speak French-Canadian French.” This is nonsense.

Now multiply this problem a hundred-fold, or a thousand-fold because you don’t have any language policy on your instance. The problem was already untractable. It won’t become a simple problem by adding more languages. I’ve seen small instances implode due to the weight of moderation. At some point, the person who ran the instance said “fuck this!” and closed the doors. These folks were not even thinking about the issue of language.

I’m not a fan of restricting languages, but I cannot fault Fosstodon for having an English-only policy. That’s for the reasons I gave above. However, I can fault Fosstodon for the way in which they delivered their message. It stinks. Pretty badly. They explain their stance here.

Unfortunately — despite being spread over three continents — the team at Fosstodon are all native English speakers.

If you don’t see the problem here, you must be a native speaker of English. This sentence may have been true when it was written. However, what on Earth does it matter whether the team are native speakers or they learned English as a secondary language? If you think that I cannot moderate in English because it is not my mother tongue, you, my “friend”, are sorely mistaken. I think I’ve proved it already. If you need more proof, well, go fuck yourself. I converse daily with people who do not have English as their mother tongue. I’m sure a portion of them would make good moderators.

The rest of their explanation is okay, but it really reads like it was written by technologists, who have not spent any actual time thinking about the issue of what translation means, when you get into the nitty-gritty of it.

Well, I hope this has been enlightening.






3 responses to “Moderation In Foreign Languages Is MUCH Harder Than You Think”

  1. Fish Id Wardrobe Avatar

    @yourautisticlife ^^^^Enlightening on the problems of multi-language moderation.

  2. Nick G Avatar

    @yourautisticlife I'm guessing that insult would translate approximately to "you to me are like a dog" though it sounds much milder in English than you described.

  3. Sally Strange Avatar

    @yourautisticlife people aren't saying their policy is bad because they're unaware of the fosstodon explanation for it. They're saying it's bad because it actually creates more opportunities for harassment than it removes.

Leave a Reply

Your email address will not be published. Required fields are marked *