Hello everyone!
In the Brazilian Portuguese group every time there are many jobs I notice some half done jobs with just the most simple sentences translated. This practice leaves a half job with only the most complex sentences, usually highly idiomatic and time demanding. The translator who takes the job is not going to take advantage of the simpler sentences and while he is translating these jobs, other simpler jobs are being taken in the same fashion and it is going to repeat on and on.
Would it be interesting to have a penalty for this kind of behavior? For example: if the translator does not send at least 80% of the translation he/she will not be able to pick up another job for at least one hour.
I guess it would not hurt the speed we work, maybe even improve it, and would avoid a few people taking advantage of simpler sentences like they are doing now.
Does it sound like a good thing?
31件のコメント
This isn't regarding the language pair the OP is mentioning, but I've seen this behavior in my own pair, JA-EN. Currently, the Hotpepper jobs (for restaurant listings & sample menus on a gourmet website) have someone going around and filling in all the menu prices, and the menu prices only. I understand that this is an option, and there's really nothing to be done about it, but a) they're not doing a good job even with that—"700JPY (tax excluded)" is a clumsy & wordy way of saying "700JPY (plus tax)"—and b) they occasionally try to translate a few of the shorter pieces of promo copy in a stilted and lackluster manner; it's clear that they don't have any pride in the job they're doing in their rush to pick off the easy jobs.
Now, there are group jobs where I think picking-and-choosing is inevitable: the DMM group jobs in JA-EN, which consist of titles & promo text for adult videos, often involve a spectrum of niche vocabulary; it'd be very difficult for those jobs to get done if one person had to translate every single title. (They're usually done by several translators picking away at them.) In cases where it's just one lackluster translator running ahead & picking the low-hanging fruit because, frankly, that seems to be all they can reach—well, like I said, I don't think there's anything that can be done, but as someone who cares about quality, it is frustrating.
@C Nascimento: We also don't like this behavior and regularly warn translators who do this frequently. We encourage you to write in to Support with the translator's ID number so we can send the appropriate warning and continue to watch their behavior.
As Gengo follows a crowdsourced business model, translators should be able to choose which translations they want to work on. However, doing a half-hearted translation is not what we support and unfair to other translators.
We have also been thinking along the lines of a penalty for this sort of behavior, and will continue to work on it until implementation later this year or early next year.
Thank you for your reply, Megan!
It's really good to know you are taking care of situations like this one. :)
I just wanted to share a few thoughts about this issue, because while I can definitely understand the frustrations that have been put forward here, I still think there are merits to the group collection system and declining unfinished collections. First of all, I totally agree that running through a collection and cherry picking easy tasks without any concern for overall quality is problematic. However, I don't feel that translators should be punished for working on a few jobs to the best of their ability and then declining the rest of the collection. Blenheim mentioned the Hotpepper jobs, which I think are a good example of why it's sometimes necessary to leave a collection unfinished. The Hotppepper job collections can be relatively large and typically you will have three to four restaurant descriptions to translate in one collection, all of which can vary greatly when it comes to types of food. So in one collection you may have Italian, Japanese, and Korean cuisine all one collection. Some translators might be able to take on all three, but if a translator is particularly skilled with say Italian and Japanese cuisine, I don't think they should be punished for doing those tasks and then declining the tasks that are related to Korean cuisine. This is of course just one example, but I think these Hotpepper collections are representative of any large order that Gengo splits up into collections. It might be one collection, but the content of each individual task can be drastically different from one another.
I know this isn't exactly the type of behavior that is being complained about here, but I just don't want to see a good system suffer just because a few translators are taking advantage of it. I am all for tighter restrictions that clamp down on deliberately bad translations, however, I hope whatever penalties Gengo comes up with don't punish translators for doing the type of thing that I'm talking about. I think Gengo would see a serious drop in productivity if translators are suddenly expected to translate an arbitrary percentage of a collection that may contain dozens if not hundreds of tasks that are lumped together from various sources.
Hey Evan.
I totally see your point. In my pair I have never seen Hotpepper jobs. Maybe that's why I haven't seen any situation when dividing a collection would be the best option.
Couldn't these collections be broken into smaller jobs based on cuisines, for example?
Hey C Nascimento, yeah I think they must just be Japanese related pairs, but either way I think you are right! I think dividing them up more would definitely help (although I suspect with some of these bigger job collections Gengo can't really help it since the projects are so massive). Unfortunately this is kind of a case by case situation, but in the meantime I think it would be good if we could just find a way to encourage more teamwork between the translators working on big collections.
@Evan: I really like your idea of encouraging teamwork between translators :) Suggestions are always welcome!
@Megan: Thanks! I'll definitely post some suggestions or ideas here if I can think of anything helpful.
Re: encouraging teamwork between translators on large collections, maybe Gengo can create a discussion forum for that specific collection and post it on the Comment section?
Hi.
This really happens a lot, specially when the job is a footage subtitle. Some of the parts are just wrong and split, but.. well, you can't fix them at all, because someone translated one part of the sentence wrong and you have to try to do something different. (Believe me, it happens a lot).
Damn, we work hard to get the things we don't know, do researches, ask people and some just prefer to do this kind of thing? It's just sad.
I know I've said my piece above, but I see the same behavior happening again with the current Jalan JA-EN jobs, and more translators seem to be joining in on it. Many more short jobs are being submitted with shoddy work—putting commas in the wrong places in addresses, ungrammatical descriptions such as "7 minutes walk from So-and-So Station," etc.
I've reported the behavior as Megan mentioned above, but I do feel compelled to note that the problem seems to be getting worse with these large pools of group jobs. More people seem to be joining in on the rush for easy money.
I feel the need to second Blenheim here. I didn't see the Hotpepper issue (I was busy working on other jobs) but I just finished a Jalan job which I picked up and, to my dismay, the original translator had "translated" just names of places, addresses, and other simple parts of the translation, leaving the descriptions of places blank (these are just 2-3 sentences, not that hard.) So I just did the descriptions (and found out a few errors in the names of places that had been translated but could not do anything about it since we can't edit someone else's work.)
I'm sorry to hear you think this type of behavior is getting more widespread. We DO warn any translator we find doing this, or is reported by another translator. If we have to warn them more than once, we revoke their qualifications and they can't translate with Gengo anymore.
Although we obviously don't agree with this behaviour at all, we have found that a lot of translators genuinely don't realize that they are doing something wrong and feel like they are doing their best to translate the parts that they can. But also, there are translators who are doing it intentionally to game the system.
Does anyone have any suggestions to help curb this behaviour? I know we need to better educate translators on what is/isn't acceptable in our system. What would you guys think about something like not being paid unless you complete the entire collection or something like that?
Re: Megan's suggestion "not being paid unless you complete the entire collection," what I'd like to know is - Is exceeding allotted time considered declining a job?
Last night I did a ShapeUp collection of 200 jobs, which came with instructions and style guide. My preference in working on translations with over 100 words is translating on my CAT tool for quality assurance and then copy-pasting the translations onto the collection page when it's not a file based job.
Little did I know how time consuming the copy-pasting one by one could be when a collection is broken up into over a hundred jobs - valued lesson learnt. My time ran out before I could finish pasting all of my translations.
I would have wasted hours of hard work, if I didn't submit part of the translations beforehand or wouldn't get paid unless I completed pasting the entire collection and submitted it.
PS. The jobs that I didn't get to finish pasting translations on are easy and straight forward, mostly names of states and cities.
Re: Megan,
I don't think translators should be forced to complete a full collection. Sometimes it is better to leave some jobs when you have no idea what to do. I noticed many times when this occurred someone completed the job in a few hours, so someone may find easy what someone else thought was really complex.
Well, maybe translators could have to complete 50% or 70% of a collection. That would make handpicking really more tiresome as translators will need to scan all the jobs to find the easiest ones.
Also, if the job is already more than 50% done everyone could freely translate any part of it to their best skills as it would hardly be a case of handpicking anymore.
A notification in case just a few jobs are translated would also be good so people can get aware of this kind of practice not being good.
And maybe an account lock like: "You've just accepted a collection and are about to send less than 50% of it. If you do so, you will not be able to accept a new collection for the next 15 minutes." Since this practice becomes profitable as the translators go from one job to another, 15 minutes would be really a downside.
As people have commented above, crowdsourcing is really good for Gengo so I think we should avoid any harm to it as much as possible.
@C Nascimento - Very good suggestions! All these rules could be monitored by the system and would hurt the "bad guys" while not affecting the "good guys" (and girls).
If a translator cannot complete the whole translation the best penalty would be to discount certain money percentage from the project, depending on the number of words untranslated (for example (10%, 20%, 30%, etc.). This way translators would research more instead of abandon projects when they face problems, which is an unprofessional behaviour.
I'd probably lean toward relying on more responsible translators to report the more egregious cases for now. I personally was alluding to a specific behavior in my post: rushing through a large bunch of group jobs to translate only the "easy" jobs before anyone else can get to them, and translating even those easy jobs poorly as a result of the rush. Though the number of translators who are engaging in that behavior is growing, it's still at the point where I think the bad apples can be pinpointed and warned off.
@Siaoning - I feel your pain, and you're not alone. Some feature requests had already been posted asking for a way to avoid this issue (and at the same time improve translation quality): https://support.gengo.com/entries/61191144-Export-contents-of-group-jobs-to-files and https://support.gengo.com/entries/21482182-Option-for-translators-to-download-upload-files-instead-of-having-to-use-the-web-translation-interfa - But, despite the requests and the votes, it seems Gengo is against it (for whatever reason...).
Re the "cherry picking", I think there's no easy solution to this, but I must agree with Blenheim, in that deterrence and penalties would seem the most effective way to go about it, provided Gengo has the ability to handle something like this and evaluate case by case, in order to distinguish the real abuses from cases where the translator actually had a valid reason for declining a partly translated collection.
Sometimes in fact "collections" contain lots of isolated strings without any context, and customers often aren't around to answer questions (and sometimes those collections are even auto-approved, probably meaning that the translations are automatically published via the API). I remember declining some collections myself in the past, because I felt it was simply impossible to correctly "translate" the remaining segments without an explanation from the customer (e.g. isolated words used together with multiple variables/placeholders, or bits of sentences that seemed to be meant to be attached to other sentences, with evident grammar and syntax issues...). Perhaps, declining a partially translated collection could be subordinated to specifying a particular reason, and after that, Gengo CS could be automatically notified of it (and then maybe carry out spot checks with the help of STs, if checking ALL of the cases was too much).
At any rate, if it were possible to "unify" collections in a single file/document (as suggested at the beginning), then "collections" could also be treated as a single job with regard to approval as well. You complete the translation and submit it, or you don't submit it at all... Obviously, for this to really work, customers SHOULD be around to answer questions and modify translations accordingly, if needed. There's no escaping that, and if a customer isn't prepared to do it, then they must be prepared to accept potentially inaccurate/wrong translations.
I've noticed this too on the recent Jalan jobs and I am curious as to why these translations are split into so many parts in the first place. I can understand separation to some extent on such a large project, but if each shop/location description (inc address, opening days, times etc) was grouped as one job within the collection, wouldn't that prevent this kind of behaviour (at least on the Jalan project)? Currently even the location/shop name and the description are listed as separate jobs - which could prove problematic for continuity, especially as previous translations cannot be edited.
Personally, I am totally against anything that would force translators to work on projects they don't want to work on. That is one of the main attractions of Gengo, after all. The main project I work on often has individual jobs within a group that are difficult to translate, especially for the newer translators on the project. I will often pick up those jobs that someone else has declined due to a simple lack of experience with this client and subject. Personally, I would prefer the translators declined these jobs and left them for someone more experienced rather than being forced to translate them or not get paid for the rest of the job.
I understand that in some cases cherry picking is a problem, but please don't punish everyone (and eventually the client who may receive translations that are not up to par) for the actions of a few. I much prefer that translators who engage in this practice be banned. Send an e-mail to everyone, post a notice on the dashboard to this effect, then ban anyone who blatantly cherry picks.
I know this is an old thread, but I would like to talk about something that just happened to me in the JA/EN language pair. I was working on a project and had just a few characters left to translate, when my time ran out. I missed out on collecting my payment by literally just a few seconds. I also noticed that part of the document had already been translated, but only the easy parts. Do people who do partial translations get paid for the full amount displayed? And more importantly, who is getting paid for all the work that I just did? It says that the collection is no longer available, but since I was working on it for about 3 hours, doesn't that mean no one else was working on it?
I would also like to note that the reason I didn't press "submit" right before the time ran out was because I wanted to submit a complete, good quality job. The system that Gengo has in place no (no pay at all if you exceed time limit by a few seconds) is a good way to ensure that translators submit unfinished or poor work just so they don't get shorted on what they earned.
Hi deraluce,
I am sorry that you have had this experience.
When you say that part of the document that you were working on "had already been translated", do you mean that this was a collection broken down into several jobs, some of which had already been translated and submitted by someone else, and were therefore uneditable? If that is the case, the person who submitted those jobs will get paid for those jobs only (and not the full reward shown.)
If, on the other hand, the job had been declined by a previous translator without being submitted, and the translated parts could be edited, then the person to complete and submit the job would get paid in full. Translators only get paid for the jobs that they complete and submit.
In general, job deadlines are calculated in a way that ensures that our translators will have plenty of time to provide good quality work. However, if you ever feel like you need more time, you can write to support@gengo.com to ask for an extension.
I am afraid that the collection that you were working on might have already been picked up by someone else, but I hope that this information will be useful to you in the future.
Lara
Hi Lara,
Thanks for your reply. I think the job was a collection broken into several jobs, as some sections of the job were greyed out and could not be edited. This was my first time taking on a larger project, but I took it on because 3 hours seemed like plenty of time. Then I hit a snag because one of the sections had a bunch of food words with no English translations. Cue researching in various dictionaries and cooking sites, looking for alternate kanji readings, etc. It didn't help that that section was just a list of words with no context. Lesson learned, I guess: submit a 99%-complete translation if time is about to expire.
Hi deraluce,
I think the actual lesson to be learned from this is: ask for a time extension if you need more time. You can do this in the support chat and the people there are quick to help, I never had a problem receiving an extension quickly if I needed one. So if you have a three hours job and only a quarter of an hour is left and you see that you cannot make it in time, don't rush your work and don't submit prematurely, but ask for an extension right away. No big deal.
kvstegemann, I think you're right. I didn't know if Gengo would respond fast enough if I were to ask for an extension, but they responded to my email pretty quickly after I told them what happened. So it looks like they're pretty fast with emails. I'll try using an extension in the future if I need to.
Dear Deraluce,
Trust me, they respond fast. I had technical difficulties a couple of times and didn't want time to run out, so I contacted support. Both times they got back to me within 10 minutes, if that. One time the technical issue resolved itself; with the other one, they extended my time so I could sort out the issue on my computer.
Hi there! Sorry for digging this up - I have no idea if the problem still persists, but wanted to give my 2 cents in case they end up being helpful.
Would it not be easiest if there was a system requiring you to give a valid reason for declining a collection halfway done? That way, you could check up on the reason the job was declined and penalize according to whether or not that reason is correct.
This would make things like the Hotpepper jobs Evan described possible by simply giving the reason "unfinished jobs required knowledge of -Insert topic here-", and the same people that quality check completed collections may then check a few declined collections for veracity.
This would allow me and other translators to decline a part of the collection that requires knowledge of such things as contract language in a way that can be verified.
I know there is already a set of reasons for declining jobs, but it must not be checked at the moment, or otherwise this problem wouldn't persist. There should also not be any generalized reasons, as such things are hard to objectively judge and end up being a matter of opinion. You can judge objectively if a job has a lot of scientific terminology, but judging if a job is too "complicated" would be mostly up to the person reading it and thus hard to control.
You might also want to give people a few "get out of jail free cards" to decline collections half finished when real life interferes. Otherwise the system might lead to resentment when someone is punished because he had to decline a collection due to a family emergency, when he or she was otherwise a very reliable and meticulous translator. There shouldn't be too many per month, as that would encourage people to take up jobs even if they know they need to be somewhere soon and can't possibly do all of it, but there should be a few to prevent punishing people because Murphy got his word in.
I can't think of any other way to ensure people cannot "game the system" by doing only the easy jobs, other than just flat out not allowing partial submissions. Which would be bad for all the above reasons.
Hello, this is old but was the closest I could find about declining collection. Pardon me if there is another.
Gengo has worked on and changed the system mentioned in here, now you have to finish ALL the jobs in a collection to get paid.
I'm an EN-JA standard translator, I've seen those kinds of collections left undone for Amazon or TripAdvisor, which I personally did not find stressful even though it as unbelievable anyone would mess it, because I was getting paid for how small I've worked.
I understand it was a burden for Gengo for fixing and all, but I feel it is unreasonable to not getting paid for, for example, 2 hours of work I did truthfully, by leaving a couple of unfamiliar contents that I cannot figure out. (I would avoid to submit uncertain translations in terms of customer satisfaction AND review) Besides, it takes more time for research and finish all of it (It was for TripAdvisor especially).
I have emailed this matter to support and they replied that they heard my view. Is there anyone else feel this way? and I would like to know if I'm missing anything to make this work..