09:22:10 please consider that it's not really a surprise that these translators are submitting poor quality work. They didn't contact me or the localization workgroup and they haven't used the glossaries available for the technical terms. This is different from attempting to scam tho. 09:22:55 That's why is not a good idea to fund people with no prior experience in Monero or in Translations and it's even a worse idea to fund two people in the same CCS if we cannot verify they are actually two different people. 09:23:47 If they use machine translations with some words changed (which most likely is also done in software) it is a scam IMO 09:24:16 This happened in past: a translator is funded to do some work that ends up to be of poor quality and somebody need to heavily change it during a review. The only difference is that usually it's me in PM that deal with the mess and it's not public. 09:24:36 I know that software like this exists that replaces words to avoid being a 1 to 1 translator copy. 09:25:34 Low quality work without using machine translations is not a scam, I agree with that. It is just low quality work. But that’s not the case here. 09:27:04 selsta: It is not IMO. If the translation is fluid and doesn't sound mechanic, it doesn't matter if they use a machine translator as a base. It'sactually a common technic, even weblate offers it as a standard (i chose to disable it). So, i agree it's low quality, but i wouldn't call it scam, because i don't think it was made with malicious intentions (talking about greek, dutch and French). 09:27:20 disagree 09:27:29 but that’s just my opinion P 09:27:31 :P 09:28:11 5 similar looking CCS that all suddenly use the same machine translation concept without disclosing it is not a coincidence 09:29:20 I've been dealing with this shit for 4 years and i helped countless people who faced the same problem. It's definitely poor work, but that's not really a surprise, as i explained. 09:29:25 Using machine translation as a base can be fine if it is disclosed prior to the work. 09:30:32 dEBRUYNE reviewed the Dutch and it is simply DeepL with some words changed without any crypto context. 09:30:53 No idea why we should pay for automatic translations. 09:33:38 The french translator answered me and according to them French is all machine translations with few changes. This is another work that should be rejected along with german. 09:34:16 The dutch translation need to be heavily adapted. 09:35:32 Why would you reject French and German but accept Dutch with adaptations? 09:36:20 Just trying to follow here. All languages are machine translations with a few changes. 09:36:57 https://usercontent.irccloud-cdn.com/file/xpI4D79s/Screenshot%202020-12-01%20at%2021.52.42.png 09:37:38 See here Greek for example, almost a whole block of text identical to Google translate, with a few single words changed. I know the quality of Google Translate for German and these are unusable. 09:38:00 Or here https://usercontent.irccloud-cdn.com/file/zjgl62vG/Screenshot%202020-12-01%20at%2021.47.10.png 09:39:29 As i mentioned, using machine translations as a base it's not a problem and it's a technic used by some translators. It becomes a problem if the machine translation is left as it is and the technical works not translated/localized. The dutch translator did an acceptable job and their shortcomings are related to the fact that they don't know technical words and they left the literal translation there. 09:39:56 Refusing all the work would mean revert ok translations that would also need to be tweaked 09:40:06 dEBRUYNE said he would reject the Dutch translations 09:40:14 and he is a native speaker so I would trust him here 09:41:23 I posted this before "Passionate cryptocurrency user and enthusiast." "Cypherpunk activist." 09:41:31 them not knowing better is completely unbelievable 09:41:39 or they lied 09:43:45 IMO paying 30+ XMR to low effort machine translations is just disrespectful to all the volunteers who put effort and thought into this 09:44:01 and someone else will have to fix it anyway if we accept these translations 09:45:48 Well, dEBRUYNE haven't dealt with translations from the past years. I respect his opinion, but he doesn't know the average quality of the translators and the work there is behind to make sure we have high quality work. Nothing of the usual happened because i wasn't aware of the existence of these guys. 09:46:11 I agree translations shouldn't be accepted as they are, but please don't blame the fact that they used machine trnaslators 09:46:29 the problem is that they didn't adapt and worked tjhe translations after as they should have 09:47:47 I blame the fact that they used machine translations without disclosing it, but I explained this before. 09:48:02 You **can** use machine translations if you disclose it. 09:48:13 well, that's your opinion :) 09:48:17 sure 09:48:35 i only care of having an high quality work. Don't care how they get there. 09:58:39 Reguarding the practical actions to take, i would ask the translators to review all their translations and resubmit them making sure to follow the guidelines of the localization workgroup and increase the quality of their translations. After they are done, i would ask trusted translators to take a look again (maybe giving one xmr each as a thanks for their time). If the translators are still not of high quality, we 09:58:39 decide what to do (rejecting the ccs? hire trusted translators to fix the mess? we can see later. I'll make sure these guys know how to reach me and the localization workgroup 09:58:40 luigi1111 i need an input from core about this, so i can contact the translators and let them know the decision 10:02:59 I would like to see them being able to do high quality translations first before continue funding them. 10:03:10 But core has to decide that. 10:03:45 I agree with that. They should rework the translations they already submitted 10:06:02 Side note: This whole discussion was basically offtopic here. I encourage somebody from core to join #monero-translations so that we can discuss there next time. This discussion is important for other translators as well 10:25:03 To sum up the situation for each language: Dutch: Technical terms need to be fixed and adapted, everything should be rereviewed regardless. German and French: low quality work that should be redone. Greek: I think it's in the same situation as Dutch, but i haven't asked any translator to take a look, since it wasn't considered a problem at the beginning (and we still don't know about the quality of the work, we only 10:25:03 know that they used a machine translator as a base) 10:37:02 We don’t need a native speaker to know that Google Translate will result in low quality work lol 10:38:52 https://usercontent.irccloud-cdn.com/file/FNhSv5km/Screenshot%202020-12-02%20at%2011.38.43.png 10:39:14 didn’t even bother to change a single word lol 10:39:59 And even in the cases where they changed a word, it still is a DeepL suggestion. You can click on words in DeepL to get a synonym, that is exactly what they did. 10:40:16 At this point I doubt they even speak the language 10:42:43 again not a single word changed https://usercontent.irccloud-cdn.com/file/dXoHBP7R/Screenshot%202020-12-02%20at%2011.42.19.png 10:43:14 Didn't we go through this already? 10:47:05 No, I just wanted to point out the the "changed words" are also directly from DeepL. 10:47:13 They are scammers. 10:48:55 I did not find a single sentence yet where they did any manual translations. 10:49:54 the dutch speaker is a native speaker. According to my translator, who reviewed a bunch of strings and tested them on deepl and google translate, there are some words added here and there that are not in the original text, so they didn't simply use machine translation and used a synonim, but added words to make things smoother. Sure it's not a good work regardless, but that's another story. 10:50:20 You get the added words by simply clicking on a word in DeepL None of this is what they did themselves. 10:50:59 I went through a lot of examples. Every time when a word is different it is the first suggestion when clicking on it in DeepL. You can check it yourself. 10:52:34 Do you have any proof that the Dutch is a native speaker? Or did he simply told you? 10:52:45 i can just report the opinion of a native speaker, i don't see much utility in running all strings on deepl myself, since what we care about is: high quality or low quality. We can ask another native speaker to make an analysis and let us know if the translations feel mechanic or not. 10:53:04 again, i can just report the opinion of the translator i contacted 10:54:35 You have no proofs that they are not native speakers tho. It's something hard to prove. 10:54:43 lol 10:57:59 well, i have enough. I am here trying to fix a mess shouldn't have happened in the first place. selsta, you are free to make your analysis and to laugh at me if you want, i'm happy to not deal with this anymore, after 4 years and hundred of Monero translators i dealt with, i think i know what i'm talking about. Please core let me know what to do. 11:04:37 doomero! 11:05:58 please let's continue this topic on -translations. 11:11:43 I'll be away for most of the day. I'll be back this evening, if core has something to tell me, contact me please. Things will stay frozen on weblate until i know something. /end 11:29:38 fyi i'll be afk for some hours (evening in europe) 15:53:06 atm I'm of a mind to throw them all out as gross misrepresentations of their expertise, at least. 16:14:16 is payment in question, or just whether to merge / fix / trash the outcome? 20:00:35 -xmr-pr- 1Bit-Group opened issue #1349: Add 1bit.group Anonymous exchange to the list of merchants 20:00:35 -xmr-pr- > https://github.com/monero-project/monero-site/issues/1349 21:05:38 I agree for french and german (i still don't know the quality of the greek translation), but i would ask the dutch guy to remake the work, since it's of higher quality then the others. If it's still not of high quality enough, then there aren't many choices 21:13:55 IMO core should not have any particularly worthy opinion on whether to use the poor translations somehow. Personally I'd leave this to the localization working group to decide 21:41:08 did some stuff get merged from faulty (cross) approvals? 22:22:47 Hello, just read what Ericcione said about the translations. Yeah, we had this conversation with the Monero Outreach Workgroup, this when rehrar also said some proposals looked like if a bot had done them. I had this feeling of bad quality. 22:30:40 I proposed to the MO we could search for quality reviewers, including translators, if needed, and help build the terminology and review all the translations. This has been something we have been working since the MO started, language is our daily work. 22:31:54 We could put the experience but we will need members to help out, it's a bunch of work. 22:42:16 getting quality translators is a good goal 22:44:27 luigi1111w: no. All the changes are still on the repository on weblate. 22:45:14 I also want to point out that the system we have been using always ensured high quality of the translations 22:45:56 ensuring high quality is a bit different than finding said translators 22:46:22 anyway neither here nor there 22:46:31 the issue at hand is what do we do with these 22:46:35 Yes it does, that's right. This would have to be a community effort and complement the structure there is already. 22:46:57 We have had hundreds of translators in these years. Quality is ensured by having a reviewing process, which the merged proposals bypassed 22:47:01 mmm how many are there, I'm lost in the conversation 22:47:13 (and what CCS rule changes are in order, but I think we are pretty good there, at least a good starting point) 22:47:23 lh1008[m]: I'm not sure of what are the improvements you are proposing 22:47:39 outreach is really a different topic than this 22:48:26 yeah that's why i'm confused. They have their own system that systematically rewards translator. Much different than the structure of the localization workgroup. 22:49:56 beside, i gave my opinion about how we could act. I'm just waiting for an input from core, since there are CCS proposals involved 22:51:59 Well, finding groups for example. What I usually do is look for them until we find them and start a personal approach to each and one of them. Ensure and explain that google, deepl, or whatever other automated tool is not accepted and validate with these tools when the translation is done. Yes, this has been working with the compensation model we've built, that has ensure us to find those who are really passionate and, 22:51:59 I call it love, to what is being built in Monero. 22:52:24 ErCiccione[m] can you (re)summarize what happened as far as Weblate/the current situation/what is necessary to fix/undo? 22:52:57 Filtering happens when they have to open accounts, taiga, github, use the terminal to push work. Most of them retire before even starting. 22:53:26 I'm away on a sort of vacation and haven't been able to keep up that well 22:54:06 In weblate, a spanish contributor used google and I noticed it. So in the proposal I made I invited another contributor to work with me and we managed to remove the google translation before merging. 22:54:23 I suggest finding a native speaker who wants to collaborate for each language 22:54:43 Someone who is really interested in what we're doing. 22:55:51 The native speaker should speak well english too, if not it's really hard. 22:56:08 luigi1111w: translators submitted and reviewed each other's translations, which got committed on Weblate. In the case of French and Gewrman, the translations are of very low quality and machine generated. Dutch is of low quality when it comes to technical terms, but the translations should be re-reviewed regardless to make sure they are accurate. 22:56:59 i suggested to ask at least the Dutch translators to rework their translations 22:57:27 then we will need a trusted contributor willing to take a look at the work to let us know if it's of high quality enough 22:57:49 I think reviewing each other's low quality work to get it committed is grounds for dismissal on bad faith 22:58:43 it's definitely something i wouldn't have accepted or endorsed, but i wasn't aware of this proposals until the mess was unvelied. 22:58:44 I looked up online, for the prices they are asking we could get professional technical translations. 22:59:13 well it was presented as professional in the proposals, right? 23:00:01 They mentioned previous experience in some cases, but in others did not and they simply called themselves "cypherpunks" or "passionate", not more 23:00:46 A dismissal for low quality makes sense, as i said, but i would give a chance at least to the dutch guy, who seemed the most professional of the bunch 23:02:00 maybe kicking them out to resubmit under the new rules if they want is the best bet 23:02:06 even if I hate post facto stuff 23:02:20 Regarding CCS proposals, i would put as a rule that one single translator is funded (two translators reviewing each other is agaisnt the reviewing process i set up on weblate) and must have verifiable experience and/or be somebody who contributed to translations before 23:02:53 On this website as an example https://translated.com/translation-rates 23:03:07 I think hard weighting to the latter is appropriate, unless an exceptional case comes around 23:03:10 don't know what the new rules will be, but if we require previous work on Monero none of them fit 23:03:33 "verifiable" is probably too much work in a lot of cases 23:03:34 What we did in MO was to decide to only work in 4 languages, Spanish, Portuguese, French, and Russian (mr pickles and team, they work by themselves and they don't ask for rewards). We're now working in the minimum amount of files that are needed to open a new language. I'm currently doing the list. This has taking me quite some time. 23:03:34 Monero could do the same, it has worked for us to control only 4 languages and even with that we're behind some translations. But well, we did almost 20 translations these last quarter expecting to be pushed to the site at any moment. 23:03:56 yes we would require them to submit high quality work on something monero before getting another proposal up 23:04:03 yeah. I always endorsed past translators, mostly because i knwo or i can easily verify their past works 23:04:23 *i know them 23:06:11 Erciccione's work is heavy, is what I can tell you all, we do the same in a smaller size. 23:13:37 I can tell you how i would proceed on my side, then i need to know how core wants to deal with their open ccs luigi1111w: Let them redo the work and i'll ask some translators to review the rework (maybe giving those 1-2 xmr as a token of gratitude). If the work matches the standards of the localization workgroup, good. Problem solved. If it doesn't, they won't get their proposals unlocked and i'll see with 23:13:37 translators what is more convenient to do, if just reverting their work or fixing it. 23:16:18 We would get professional technical translations with revision and quality control for less than they are asking. And they would not use machine translations. We have to decide if what they are delivering is worth our time and funds. 23:18:26 Also IMO undisclosed machine translations quite a breach of trust, considering that it is industry standard to disclose and offer reduced rates when used. The website I previously linked asks for 1/4th of the price when machine translations with manual revision are used. 23:18:30 Definitely, yes, but their proposals got approved on those terms and that should be considered. 23:27:14 lol. Just came in my mind that some would prefer to move all ccs proposals directly to funding required without vetting. This situation is a great example of why vetting is necessary. 23:27:49 yep 23:28:32 We use DeepL Pro because it's a standard tool all translators use. It's faster that way and it helps us translate better. 23:28:45 https://translate.getmonero.org/translate/getmonero/monero-site/de/?offset=1&q=deepl&sort_by=-priority%2Cposition&checksum= 23:29:16 this clearly says they are using the free version, so claiming to use DeepL pro seems inaccurate 23:32:15 yeah, as i said i would be ok with rejecting German and French for poor quality. I would give a chance to the Dutch guys. The proposal was approved with 2 translators working on it, so i guess if we let them finish they can do it together (unless core decided to change things "post facto"). But as i already said, having two translators in one ccs is a very bad idea. 23:34:02 I talked with dEBRUYNE again today and we came to the conclusion that German and Dutch are similar from the quality 23:34:17 s/we let them finish/we make them redo the work 23:34:41 both look good on first look because DeepL simply is good at translating basic texts, but once you look at the details things start to get wrong 23:35:01 like basic cryptocurrency terminology getting translated with wrong words 23:35:27 e.g. mining with Bergbau 23:36:30 or mixing with the thing a DJ would use in a club 23:41:52 Even if they fix the technical wordings the price they ask is outrageously high. 23:42:18 Getting technical terminology wrong is not as uncommon as it would appear, usually the problem is solved by having experienced translators (of the localization workgroup) reviewing and by asking translators to use the glossaries we built specifically for that purpose. We can also ask to all of them to redo the work they submitted until now. If known translators of the localization workgroup who have experience with 23:42:18 unexperienced translators submitting work consider it low quality, the CCS translators don't get their proposal unlocked. 23:43:02 the price they ask is outrageously high. <- That's not something we should deal with. Their proposal was accepted at those rates, that yes, are much higher than any other translator before them 23:44:01 But the proposal was accepted without them disclosing to use machine translation assistance. That has to be considered. 23:44:11 i usually suggest the rate to translators who open ccs proposals (most of them contact me before opening the proposal, or after), but as i said, i wasn't involved at all in evaluating these proposals and i wasn't contacted by the submitters 23:44:16 Like I said, in the industry that would cut the price significantly. 23:45:39 Yes, but we don't really follow industry standards and we never mentioned disclosing the use of machine translators 23:46:17 that doesn't mean that shouldn't be rejected as poor quality work, as i mentioned 23:48:32 but IMO we cannot say "you should have disclosed the use of machine translators" if we never mentioned that they are required to do so. The enforceable point is the low quality 23:49:46 I'm going to bed. I'll see you tomorrow for another episode of these annoying saga 23:49:49 *this 23:51:19 They are wasting our time. Just cancel the CCS and move the funds to a translation proposal by a community member with history of contributing to translations would be the easiest solution. 23:52:36 Or keep the funds for community members who fix the low quality translations.