| Pages in topic: [1 2] > | How much do you rely on MT (and CAT tools in general) for your work? Thread poster: Daniel Santana
|
|---|
Daniel Santana 
United States
Local time: 17:48
English to Spanish
+ ...
Hi, everybody! I am a newbie to the translation world. I have been trying to learn how to use Trados because I know how useful and how important CAT tools are in this industry, and I know that Trados is the most widely used CAT software. However, while watching the introductory videos, it caught my attention how whenever there was no TM match MT was used in its place, instead of manually inputting the translation. This had me wondering if this is the industry standard or if it is up to translato... See more Hi, everybody! I am a newbie to the translation world. I have been trying to learn how to use Trados because I know how useful and how important CAT tools are in this industry, and I know that Trados is the most widely used CAT software. However, while watching the introductory videos, it caught my attention how whenever there was no TM match MT was used in its place, instead of manually inputting the translation. This had me wondering if this is the industry standard or if it is up to translators to decide when to use MT. I don't see anything bad with doing it, but I feel that manually translating gives me more "command" over the text, so to speak, than just doing an MTPE. I just wanted to know how much I should rely on CAT tools and, more specifically, on MT.
PS, sorry if this is labeled under the wrong topic, as I said I am a newbie to translation and ProZ
[Edited at 2022-05-01 17:01 GMT] ▲ Collapse
| | | | Stepan Konev
Russian Federation
Local time: 01:48
English to Russian
The CAT tool you mentioned above is not the only tool in the market. There are plenty of them.
Also MT is not the only feature of a CAT tool. You can disable it at any time if you don't want to rely on it.
| | | | Daniel Santana
United States
Local time: 17:48
English to Spanish
+ ...
TOPIC STARTER | do you use MT often? | May 1, 2022 |
I know it is not the only CAT tool, but according to what I have read, Trados is the most widely used; I might be wrong, of course. And thanks for your reply! I did not know I could turn that option off . Although the question was aimed more at knowing how other translators use MT. Do you use MT often, never, or just on certain occasions? I just want to familiarize myself with the most common way to use this tool.
| | | | | Trados is on the wane, and MT is controversial | May 1, 2022 |
Trados used to be the most popular CAT tool for many years, but its market share seems to have substantially dropped lately. It may differ from one market to another, but I mostly see mentions of MemoQ, Wordfast being the second. However, most CAT tools of today can interoperate fairly well, so a file exported from Trados can be easily handled by MemoQ, and vice versa.
The use of MT is a highly controversial topic. Some agencies use it as a standard, and some strictly prohibit any M... See more Trados used to be the most popular CAT tool for many years, but its market share seems to have substantially dropped lately. It may differ from one market to another, but I mostly see mentions of MemoQ, Wordfast being the second. However, most CAT tools of today can interoperate fairly well, so a file exported from Trados can be easily handled by MemoQ, and vice versa.
The use of MT is a highly controversial topic. Some agencies use it as a standard, and some strictly prohibit any MT and blacklist the translators who use it. Personally, I never use MT for professional work because it takes me longer to fix the MT output to meet my personal quality criteria than to translate the same text from scratch. However, I may use MT for my own needs (to translate from languages I know at an amateur level only), or for unpaid translations of simple texts (e.g. translating news articles on the war in Ukraine for my friends - but even then I will edit the output). Based on similar feedback from experienced colleagues, I am firmly convinced that a translator who finds MT to be a time-saving tool for professional work is not a qualified professional in the given language pair.
[Edited at 2022-05-01 21:28 GMT] ▲ Collapse
| | |
|
|
|
expressisverbis
Portugal
Local time: 23:48
Member (2015)
English to Portuguese
+ ...
| I agree, but... | May 1, 2022 |
Anton Konashenok wrote:
Trados used to be the most popular CAT tool for many years, but its market share seems to have substantially dropped lately. It may differ from one market to another, but I mostly see mentions of MemoQ, Wordfast being the second. However, most CAT tools of today can interoperate fairly well, so a file exported from Trados can be easily handled by MemoQ, and vice versa.
The use of MT is a highly controversial topic. Some agencies use it as a standard, and some strictly prohibit any MT and blacklist the translators who use it. Personally, I never use MT for professional work because it takes me longer to fix the MT output to meet my personal quality criteria than to translate the same text from scratch. However, I may use MT for my own needs (to translate from languages I know at an amateur level only), or for unpaid translations of simple texts (e.g. translating news articles on the war in Ukraine for my friends - but even then I will edit the output). Based on similar feedback from experienced colleagues, I am firmly convinced that a translator who finds MT to be a time-saving tool for professional work is not a qualified professional in the given language pair.
[Edited at 2022-05-01 21:28 GMT]
I agree with you, except that last part.
MT can save you time depending on the text you are translating. Also, I use it in a reasonable and wisely way, but that doesn't make me a lousy professional in my language pairs.
I translate exclusively to European Portuguese, and I know fairly well my native language to control machine translation.
I wouldn't dare in any case to translate, for example, a medical device or a novel with MT. In transcreation no way either.
If it is so bad like you say, I wonder why MT is one of the main subjects of University courses related to translation.
Please note I am not advocating MT, I speak for myself and perhaps on behalf of other colleagues, because considering MT a CAT tool's feature that can save you time in certain texts/fields doesn't mean you aren't qualified, and the proof is that it is taught in reputable Universities.
| | | | | Sorry, can't agree | May 2, 2022 |
expressisverbis wrote:
I translate exclusively to European Portuguese, and I know fairly well my native language to control machine translation.
Unfortunately, checking the output just from the standpoint of one's native language isn't enough. The biggest problem with the current generation of MT engines is that they may create an output that flows deceptively well yet contains totally counterintuitive errors. For this reason, editing MT output requires considerably more attention and time than editing professional human translation. MT can compete in quality with mediocre students, but not with real professionals. An MT engine is inherently unable to understand anything, it can only fake the understanding (just like a mediocre student desperately trying to pass an exam). An MT engine cannot conduct terminological research on its own. An MT engine has no concept of diligence, nor will it admit its ignorance of a certain expression or fragment, it will blindly insert the best match without giving any warnings. Sometimes it will omit a fragment altogether (I see it fairly often).
If it is so bad like you say, I wonder why MT is one of the main subjects of University courses related to translation.
Probably because there is a demand for it. Why is there a demand? Because 70-80% of our clients have no understanding of quality. Because people are greedy and gullible, and blindly believe in the new technology evangelised by the IT industry.
MT (...) can save you time in certain texts/field
Yes, it can, but these are the simplest kinds of texts we encounter: news articles, personal documents, basic contracts and other texts that can be entrusted to fresh graduates who were explicitly taught the basic tricks of the trade to translate these texts.
| | | | Stepan Konev
Russian Federation
Local time: 01:48
English to Russian
| MT is just a tool | May 2, 2022 |
If you can't use a tool properly, it doesn't mean that the tool is bad or inappropriate. It only means that you are not skilled enough to use it. The same applies to MT as a tool. When the era of CAT tools just began, thousands of translators complained that CAT tools slowed their work down. MT is just a type assistant engine that automatically types words for you based on your source text. Nothing more than that. It is not others who are greedy, blind and have no understanding but it's you who ... See more If you can't use a tool properly, it doesn't mean that the tool is bad or inappropriate. It only means that you are not skilled enough to use it. The same applies to MT as a tool. When the era of CAT tools just began, thousands of translators complained that CAT tools slowed their work down. MT is just a type assistant engine that automatically types words for you based on your source text. Nothing more than that. It is not others who are greedy, blind and have no understanding but it's you who can't use the tool.
[Edited at 2022-05-02 02:04 GMT] ▲ Collapse
| | | | jyuan_us
United States
Local time: 18:48
Member (2005)
English to Chinese
+ ...
| Two different things involved | May 2, 2022 |
You can ask how much you would reply on MT, but the question of how much you would "rely on" CAT tools hardly make any sense. You either use a CAT tool on every project, or use no CAT tool at all.
| | |
|
|
|
| CAT tools yes, MT no | May 2, 2022 |
I use CAT tools for everything, and MT for nothing.
A lot of translators in my language combination rely on MT for everything, and it shows. Economic considerations aside, I find it sad that they trust machines more than their own brains.
Philippe
| | | | Samuel Murray
Netherlands
Local time: 00:48
Member (2006)
English to Afrikaans
+ ...
Daniel Santana wrote:
While watching the introductory videos, it caught my attention how whenever there was no TM match, MT was used in its place, instead of manually inputting the translation.
For offline CAT tools, this is unusual. I strongly believe that authors of such videos do the industry (and new translators) a disservice if they use machine translation as a normal part of the video. Although many translators do use machine translation, it is not yet the standard way of using offline CAT tools. In most offline CAT tools, setting up the machine translation components takes a bit of tinkering, so I imagine most users of CAT tools do not actually use machine translation.
Added: In Trados, it is now easier than ever to enable machine translation:
https://www.youtube.com/watch?v=AhP-Q1m9gyk
If I understand correctly, Trados' NMT engine (Language Weaver) offers 6 million characters per year included in the purchase price. Language Weaver supports about 60 languages.
On the other hand, many online CAT tools now come with machine translation built in. It is fast becoming the norm to use machine translation in online CAT tools. In the online CAT tools that I use, it is sometimes not even possible to disable the machine translation.
By the way, some video authors use an automated way of typing the translation to speed things up in the video, but that's just a trick of the video: it's a human translation that just appears on the screen very quickly.
I feel that manually translating gives me more "command" over the text, so to speak, than just doing an MTPE.
Yes.
[Edited at 2022-05-02 08:28 GMT]
| | | | | I use CAT but not MT | May 2, 2022 |
Ultimately, MT is based on millions of texts translated using CATs, and the accumulated translation memories. It is not as simple as that, but that is how it all began.
I know some colleagues can reap advantages from MT, but I find it a distraction, and I am not alone. I find it much quicker and easier to translate from scratch where there is no suggestion from the translation memory (TM).
On the other handl it takes years to build up a TM, and if beginners can find useful su... See more Ultimately, MT is based on millions of texts translated using CATs, and the accumulated translation memories. It is not as simple as that, but that is how it all began.
I know some colleagues can reap advantages from MT, but I find it a distraction, and I am not alone. I find it much quicker and easier to translate from scratch where there is no suggestion from the translation memory (TM).
On the other handl it takes years to build up a TM, and if beginners can find useful suggestions from MT, then there is every reason to take advantage of them.
My translation memories help with consistency when working in the same subject field, or on repeat jobs for a regular client. MT was not an option way back when I started using CATs, but my TMs include imports from agencies and generally reliable colleagues - in Trados I can see the initials of the translator who first created each segment. Of course, the great majority are my own (CA or CRA). I also know which subject areas particular colleagues are experts in, and they can be enormously helpful.
I use Trados especially for what is called the concordance - finding terminology in context - and Multiterm - the glossary database. The MT link is disabled, and I never use MT professionally, although it is handy for languages I cannot read myself.
In my professional work, the suggestions from MT are usually strictly correct, but they are not always good English. They reflect the source-language syntax, and are often very literal. I am paid to write an idiomatic text - my motto is ´translating the message, not just the words´.
With CAT suggestions, I have already adjusted the syntax and found an idiomatic suggestion, so there is a reasonable chance that it can be used again. I still check when it appears, and when I revise and proofread the translation.
It probably varies a lot from language to language, and from one subject area to another how useful MT can be. Where the subject includes a lot of standard formulations - sometimes called ´boilerplate´ - which are translated into equivalent standard formulations in the target language, then the MT algorithms, based on statistics, will find them. Where these standard phrases exist and are expected, they may be difficult to formulate from scratch, and MT is a help.
All the same, complex legal or technical sentences have to be carefully analysed and checked against the source. If they are really standard, I already have many in my TM, checked by me. If not, I still prefer to work them out myself.
In marketing, where the language is freer and word-plays are frequent, MT is out of its depth - and there is a serious risk of mistranslation. Or smart slogans in one language fall flat in another, and have to be recreated by a human. ▲ Collapse
| | | | | OPUS-CAT MT supports Afrikaans | May 2, 2022 |
Samuel Murray wrote: Daniel Santana wrote:
While watching the introductory videos, it caught my attention how whenever there was no TM match, MT was used in its place, instead of manually inputting the translation. For offline CAT tools, this is unusual. I feel that manually translating gives me more "command" over the text, so to speak, than just doing an MTPE. Yes.
If I translated HTML files in OmegaT and I have good TM with segments splitted on sentences, OmegaT do not segments some paragraph to sentences. It is faster use MT and compare two, three or four sentences in Fuzzy translation pane than copy-past them individually and so translate the paragraph.
@Samuel
OPUS-CAT MT supports Afrikaans
All bilingual models are for one direction language pair. There is no third pivot language.
Target languages: Afrikaans, 13 models
Source languages
German
English
Spanish
Finnish
French
Swedish
German
English
English
Esperanta
Dutch
Russian
Spanish
Model name
opus-2020-01-20
opus-2019-12-18
opus-2020-01-16
opus-2020-01-08
opus-2020-01-og
opus-2020-01-16
opus-2021-02-19
opus+bt-2021-03-07
opus-2021-02-19
opus-2021-02-18
opus-2020-06-17
opus-2021-02-18
opus-2021-02-19
Source language: Afrikaans, 13 models
Target languages = a same as for a direction to Afrikaans.
Model name: similar to names for a direction to Afrikaans.
--
Using is local, private, without web services and fees.
Milan
| | |
|
|
|
expressisverbis
Portugal
Local time: 23:48
Member (2015)
English to Portuguese
+ ...
| Can you explain it, please? | May 2, 2022 |
Anton Konashenok wrote: expressisverbis wrote:
I translate exclusively to European Portuguese, and I know fairly well my native language to control machine translation.
Unfortunately, checking the output just from the standpoint of one's native language isn't enough.
MT is one thing and I can agree with you in many aspects, and being native in a certain language (PT-PT in this case) is something else quite different, even in MT related-context.
If you master your own language you can control those errors delivered by MT, and it's here that our brains, education and culture work.
| | | | | @expressisverbis | May 2, 2022 |
expressisverbis wrote:
If you master your own language you can control those errors delivered by MT, and it's here that our brains, education and culture work.
I mean that today's MT engines sometimes create an output that flows perfectly and looks immaculate in the target language, but when you carefully compare that output to the source, you find it to be completely wrong.
| | | | expressisverbis
Portugal
Local time: 23:48
Member (2015)
English to Portuguese
+ ...
| Can look immaculate, but not to a keen eye | May 2, 2022 |
Anton Konashenok wrote: expressisverbis wrote:
If you master your own language you can control those errors delivered by MT, and it's here that our brains, education and culture work. I mean that today's MT engines sometimes create an output that flows perfectly and looks immaculate in the target language, but when you carefully compare that output to the source, you find it to be completely wrong.
That's why - and I think I am going to answer to the OP's question now - I rely much more on my translation memories (TMs), glossaries, termbases and dictionaries than in machine translation of a CAT tool. They are by far the most important features.
There are people who misuse (or even abuse) MT, and many times we are asked to make the revision of a text (presumably as a human translation) and we see exactly what you say above.
Fortunately, a professional and responsible translator has a sharp eye for detail that many don't have (in particular the fresh translators who are just getting started), and can spot it immediately.
[Edited at 2022-05-02 18:20 GMT]
[Edited at 2022-05-02 21:26 GMT]
| | | | | Pages in topic: [1 2] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » How much do you rely on MT (and CAT tools in general) for your work? | 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 » |
| | Wordfast Pro | Translation Memory Software for Any Platform
Exclusive discount for ProZ.com users!
Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value
Buy now! » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
