Poll: What do your clients think about neural machine translation? Thread poster: ProZ.com Staff
|
This forum topic is for the discussion of the poll question "What do your clients think about neural machine translation?".
View the poll results »
| | | neilmac Spain Local time: 04:30 Spanish to English + ... Don't know, never discussed it | Feb 14, 2019 |
Am I the only one who feels this and yesterday's poll give a slight whiff of stealth marketing? As though someone, somewhere was trying to promote NMT through awareness-raising actions like this? Or am I just being paranoid? | | | Marjolein Snippe Netherlands Local time: 04:30 Member (2012) English to Dutch + ... varies per client | Feb 14, 2019 |
"My clients" are a diverse bunch. Some have indicated they are exploring into this, some may have started using it. So far, I have not noticed an effect on the work I receive. | | | One client is introducing it system-wide | Feb 14, 2019 |
I'm very nervous about this. I commented in yesterday's poll about the problems that I"m already finding. The output "sounds" good, but there are poison cookies--passages that look OK but are dead wrong. Not to mention that the sources they draw on don't all use the same terminology or standards (the sources are a mix of UK, Caribbean, US, and UN English from agencies across the UN system). IMHO, it's a royal mess.
[Edited at 2019-02-14 09:04 GMT] | |
|
|
neilmac Spain Local time: 04:30 Spanish to English + ...
As I see it, the problem with MT in general is when it's touted as a panacea, like some kind of magic wand that always gets things right. For example, one of my clients notified me last week that I would be receiving several pages of units of measure to translate ES-EN. As a guide or aid, they sent me an Excel file which had many of the terms already translated in parallel columns. As an experiment, I took a chunk of about 300 words and stuck it into Deepl to see what would happen. ... See more As I see it, the problem with MT in general is when it's touted as a panacea, like some kind of magic wand that always gets things right. For example, one of my clients notified me last week that I would be receiving several pages of units of measure to translate ES-EN. As a guide or aid, they sent me an Excel file which had many of the terms already translated in parallel columns. As an experiment, I took a chunk of about 300 words and stuck it into Deepl to see what would happen. It translated the lot pretty much instantaneously, but when I looked over it, I found a few typical MT mistakes, usually related with polysemy. I think the problem arises when users just accept the output of systems like this as gospel and either don't bother to check carefully to ensure that the MT output is indeed correct, or when they lack sufficient knowledge of source and target language to be able to do so thoroughly. Perhaps I should add that the errors in this case seemed to come from the Spanish text I was sent to translate having been translated from an original English file. So we get things like "table" translated as "mesa" instead of "tabla"… and many other howlers, some more amusing than others.
[Edited at 2019-02-14 11:11 GMT] ▲ Collapse | | |
I don't know in most cases, but I generally do not work with clients who actively use it. I do know that one client is watching developments, and they are interested in using MT and NMT where appropriate, but they are investing in quality - and they are staying with human translators and CAT tools for the time being. | | | Mario Freitas Brazil Local time: 23:30 Member (2014) English to Portuguese + ... They don't like it | Feb 14, 2019 |
Most of my clients will warn the translator when they realize MT or neural was used (which is pretty easy to identify), and request them not to use them anymore. If the translator insists, they'll likely be excluded from the team or receive only small and simple jobs like apps, games, and others where MT may be acceptable. I have several clients that warned me not to use MT BEFORE they sent me my first job, and felt relieved when I assured them I don't. | | | One of my clients apparently thinks very highly of it ... | Feb 14, 2019 |
... given the fact that they sent a whole bunch of Google translations to me yesterday to be checked at my regular proofreading fee. After I had found out where those terrible translations came from, I explained the difference between proofreading and PEMT (and why I don't offer the latter service). Needless to say I declined the assignment. | |
|
|
Tansy Tazewell Germany Local time: 04:30 Member (2008) German to English
I have a direct client who has discovered NMT and keeps sending me little bits and pieces they've put through it to check. There's often nothing whatsoever wrong with the translation and nothing for me to correct, whereas in the past they'd send me bits and pieces they'd translated themselves - and there was always plenty to correct. It's worrying. | | | Ricki Farn Germany Local time: 04:30 English to German
They are not interested in NMT because they can have me. Cue Joan Baez's voice inside my head: "Why go looking for lilacs When you're lying in a bed of roses?" *sorry* *does impression of Homer Simpson disappearing backwards into hedge* | | | Never discussed it | Feb 14, 2019 |
It never came up for discussion. | | | How could a client... | Feb 15, 2019 |
know anything about such matters, or care? I am certain that there are new trends in dentistry technique, but I wouldn't know anything about that as my dentist's client. Why would our client try to find out about such things? | |
|
|
Mario Freitas Brazil Local time: 23:30 Member (2014) English to Portuguese + ...
Maxi Schwarz wrote: know anything about such matters, or care? I am certain that there are new trends in dentistry technique, but I wouldn't know anything about that as my dentist's client. Why would our client try to find out about such things? Experienced clients know well they should avoid translators who use MT and NMT in their work. And that's what they do, actually. | | | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Poll: What do your clients think about neural machine translation? Anycount & Translation Office 3000 | Translation Office 3000
Translation Office 3000 is an advanced accounting tool for freelance translators and small agencies. TO3000 easily and seamlessly integrates with the business life of professional freelance translators.
More info » |
| CafeTran Espresso | You've never met a CAT tool this clever!
Translate faster & easier, using a sophisticated CAT tool built by a translator / developer.
Accept jobs from clients who use Trados, MemoQ, Wordfast & major CAT tools.
Download and start using CafeTran Espresso -- for free
Buy now! » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |