This conversation should be in -translations, but i don't think anybody from core is there, so i'm writing here, since we started the conversation here.

The situation is more complex than i had hoped

they didn't only suggest and confirm their own translations, they also accepted translations of other people.

That means we have potentially legit translations submitted that got accepted and mixed with the non legit ones.

I have to take a look at each language for each project involved (GUI, CLI, monero-site), but doesn't look good. i cannot roll back, because we would lose legit translations; i cannot simply revert commits because they also confirmed what looks to be legit translations and we would lose them.

That's an annoying potato

i'll post an update later on

Is losing legit ones a problem since people who made them can presumably submit them again ?

It is. Because they did a mix of confirming their own and confirming other people's translations. We have to track every change one by one and check which one was from a reliable contributor and got confirmed and which was from the scammers and got confirmed as well. Weblate has no own tools to revert changes made by malicious users.

i'm still checking and writing stuff down as i speak, i'll post something here soon

So. I locked their accounts. in 2 of the 4 proposals they started translating before it was merged. In one case they started right after and in the last case they had not started yet (Italian). In only one case one of the user simply accepted the translations made by the other guy, in all other cases they were smarter and confirmed other people's suggestions or spreaded the work between each other to make it look

legit.

i cannot get much info about the users but they have all used linux and the same version of firefox or/and chrome

I can manually revert all their strings. it will be boring, it will take a bit of time, but it's not a huge deal. I'm thinking about what to do with the legit suggestions that got accepted by the scammers.

I may ask to trusted members of the localization workgroup to take a look and rewrite their suggestions or review the accepted ones.

This could take some time, what does core think about using a small part of the general fund to give a small reward to the translators who will be willing to review and fix? I have to take a look at how many strings need to be checked by translators first (i'll know when i'll start reverting), but if it's a good amount of time 1 or 2 XMR could make easier to find contributors

luigi1111 rehrar ^

has anybody had any contact with the translators/scammers yet? I know they wrote in -community today, but i joined only now and i have no voice there yet

for the records, the total affected strings are 891, spreaded mostly across GUI and getmonero

it sucks that there's no easy way to revert a bunch of malicious changes

does weblate have a bounty program where maybe we could give the developers a couple of moneroj to add that?

(for when it happens next time)

binaryFate also. I'm ok with paying a nominal amount

does weblate have a bounty program where maybe we could give the developers a couple of moneroj to add that? -> asymptotically: actually, yes. WeblateOrg/weblate #3040. They are on bountysource

You can put bounties on issues to make the dev work on them

Alright. i can start reverting the strings, but maybe we should have a word with these people first

rehrar i see you tried to contact them, did they answer?

.merge+ 1260

Added

So, at least one of the people claim they use deepl and then fixed the generated translation. That doesn't seem to be the case, since the translations seem to be copy pasted from deepl.

"We were thinking that once the website is online we would fix everything that is out of context." <- this one ends it for me

luigi1111 rehrar: did core take a decision about what to do with the suspended CCS? i will start reverting once i know something

one of the translators for french deny they used deepl or another machine translator. I just checked a random string and doesn't seem to be machine generated. selsta, do you have some examples of machine-generated translations for french?

The strings for french i'm looking at don't seem to be generated by deepl. Did anybody else run the strings on deepl? rehrar maybe?

Can be that they are actually a group of people and some of them are simply morons

French and Dutch seem to be legit (at least not generated using with deepl or google translate), but some Dutch translations look weird and i'm not sure they make sense. I think we need some known native speaker to take a look and give an opinion.

if somebody have a trusted french translator, please let me know. i'll contact the contributors i know

According to a trusted dutch translator the Dutch translations seem to be legit. Maybe done with a machine translator first, but then corrected and adapted. Which wouldn't be an issue if the translation is good (and that seems to be the case)

I also asked a trusted german speaker to review the strings. They are definitely generated by deepl and then tweaked here and there. They also ignored all the instructions, glossaries and translation memory. So yeah, their work should be rejected.

i'm already in contact with a french translator that will take a look soon and will let me know

Can look again, but both French and Dutch were definitely DeepL

ErCiccione[m]: usercontent.irccloud-cdn.com/file/i…ot%202020-12-01%20at%2018.38.20.png

this is 1 to 1 DeepL translation

not a single character is different

Yes. I think some of them used deepl or other machine translations as a base, and then tweaked them if necessary. The dutch translator i asked to take a look made a deep check and found some translations that a machine translation would have missed or translated different

I did not find a single translation that was deepl with 1 or 2 words changed something

we can ask another native speaker to take a look if we want another opinion

I gave the Translator this link: translate.getmonero.org/translate/g…t_by=-priority%2Cposition&checksum=

This what they wrote me just now after another check: "just did a google translate check , google actually did build the sentence better then our guy. I'm pretty sure, it's not machine translated"

Dutch is also DeepL

not google translator

so he should check with DeepL and not Google Translator

yeah. They tried first deepl and now they are checking google translator

With french, it is always deeplL with sometimes 1 or 2 words changed

so they can’t claim they did not use DeppL

So the translator should at least be honest about it

they probably did use a machine translator at first or some other tools and then tweaked it to make it smooth. I'm still waiting for the opinion of the trusted french translator

so they can’t claim they did not use DeppL <- some of them confirmed they used deepl as base, somebody else used their local tools

again such a long text with not a single word different from DeepL

We don't know about french yet. I only got feedbacks for german and dutch at the moment

undecided what to do yet. if they are clearly bad faith we'll delete

all deeplL translated with triptych manually getting replaced from translation

either way new rules are in order

they all ur using machine translations and never disclosed that

Luigi1111: i'll post updates when i know something. New rules will be appreciated.

Selsta: yes they are, but that itself is not a problem. But german for example is sloppy at best if we assume good faith.

The german one is a straight up scam anyway.

Talking about the other languages.

Yeah

Anyway, IMO it would be BS to pay for machine translations if it does not get disclosed ahead. A lot of contributors that don’t get paid put way more effort into the translations.

And Dutch, French and also Greek are clearly machine translated with 1 or 2 words changed here and there.

thanks for working on straightening this out. we'll get the new rule ironed out before accepting any more translation proposals

selsta: imo it doesn't matter how the translation is done, as long as it is of high quality. At least one of them use good tools locally, so at least some of them know what they are doing.

Who used good tools locally?

luigi1111: ok. Let me know if you want my opini

on on somethinf

selsta: rodrigoklx, one of the two dutch translators

I'm away till the weekend so might not get done right away. but anyway the translations in ideas will not get moved until after

ErCiccione[m]: I did talk with the greek guy and the Italian guy. Just a moment let me get to a computer and I'll say what I found. It's a deeper mystery.

ok. Sorry for the delay.

So basically, the greek guy says he doesn't know any of the other guys except for the Italian guy (staff guy). The others he doesn't know. After speaking with staff a bit, I think he's somewhat likely to be legit.

Did he admit to using Google translate?

greek guy? no. Italian guy didn't start translating yet.

He clearly does use Google Translate with one or two words changed here and there.

Though because of the abundance of Italian speakers we have, including in this room, Italian seemed like one of the least likely to sneak through

the Greek guy?

yes

It's getting too complicated for me to remember who did use what and who didn't. We need a spreadsheet. :P Or maybe just something here, like:

Greek: Google Translate + small changes

Italian: Not started

German: DeepL. Complete trash.

Dutch and French are DeepL with some words changed too

Dutch: Machine translated + small changes

French: DeepL + small changes

so really, none of these are legit then, huh?

rehrar: I think - for the Italian - he used DeepL too. There're some weird words that are totally out of context in past translations work he posted on Gitlab.

SerHack: great thinking. You looked at his prior work for the other orgs?

Yeesh. I wonder if we should let those people know.

rehrar: yes.

these people are definitely all connected. A group of people that have decided to band together and go around to different projects or something

extract as much value as possible, and hop away

luigi1111: As far as I'm concerned, I'm ok with paying pallpark discussed (1-2 XMR) to volunteers who can sanitize the translation mess

From General Fund.

cool

I was wondering how I can help. Yet, i as a volunteer i prefer not to he paid. Is this a good first issue to jump into?

* I was wondering how I can help. Yet, i as a volunteer i prefer not to be paid. Is this a good first issue to jump into?

jump into clarkkozak[m], jump into

how can I help? Some like some copy and pasting previous git commits, no?