Translation agencies bad practices (and how machine translation can help)

This is a copy of our posting in the LinkedIn group “Translation agencies bad practices”.

The question is “How much a translation agency pays on average to translators?” – disregarding from the start that there are staff and freelance translators with quite different aims and that one translator may be so at an organization because of personal believe (Pangeanic is a member/sponsor of Translators Without Borders, for example, where professional freelance linguists and volunteers are required but whose presence is still needed). I’m not one for wanting medals on my jacket, but we like to give back to society at Pangeanic, and thus we also support Medicin Sans Frontiers and, personally Amnesty International and UNHCR.

I am an engineer who turned into linguistics and by chances of life, began a career in translation. I tend to compare the feedback and the opinions of translators with what happens in other professions. It is a horror story. Entry level is quite low and any bilingual thinks he/she is qualified as a translator. Tools come later. Most have no idea of formatting and know practically nothing about the technology they are running on their PCs. Running a translation agency I have dealt with cases of translators not using spell checkers (and not knowing how to use them!), not using (or knowing about) coherence tools like XBench or QA Distiller… even from full-fee professionals. Practically none have heard of ISO standards.

This is the quote from our entry. I encourage all translators to

a) become aware and expert with CAT tools;

b) establish a quality, self-checking routine;

c ) become interested in the technical developments of the industry, from CAT tools (not everything ends with Trados, there’s MemoQ, Swordfish, OmegaT, tools that do a very good job for a very reasonable price). Distrust any plumber who comes to fix a leak in your house empty-handed but saying he knows a lot about water…

“The problem here is the name and the aim of the group itself. I find no other “electricians’ bad practices” or “how much does it cost to hire a plumber”. Our profession is at a time of technological change with the advent of machine translation and user empowerment. First, translators should leave their attitude of being “creative re-creationists” unless they translate works of literature (which is one of the worst paid activities in translation, if we count by the word). I have been a freelance translator, run my own LSP and developed MT software, so believe you me I am not biased at all. I understand the will and the need for a one-person company to make a living, the pressures of a translation agency having to deal with 20 languages and unreasonable deadlines and fees and, above, the help to automate. Lastly, something that I have learnt is that translators loathe and despite machine translation and their new role as post-editors (just like workers destroyed machines in the 19th-century industrial revolution) but most are not ashamed of a “bottom up” approach when it comes to using GT to boost their productivity (see – it is in Spanish but highly enlightening and you can use Google Translate to get the gist).

It is an opinion of a freelancer who has worked as a staff translator for 4 years. Freelance translators do not like being “told” they have to post-edit, although the practical change is minimum – and they can put their precious neurons to a better use, for example, adding feedback to improve technology with better translations. No other freelance profession (journalist, electrician, car mechanic, plumber, even lawyers!) would resist a tool that enhances productivity/turnover…. even if it meant lower rates but more work…. The issue is that there are not enough qualified translators around and too much content to be translated. How to tackle that and not if I’m paid 1 cent more than the average is the real point.

4 thoughts on “Translation agencies bad practices (and how machine translation can help)

  1. financialtranslator

    Hello, Manuel. I must confess I find this post and the reply to the LinkedIn forum baffling. First of all, no one mentioned machine translation until you joined the fray. I think the issue discussed in the forum was creating a database of rates paid by agencies. If this generates anxiety (“I find no other “electricians’ bad practices” or “how much does it cost to hire a plumber”) could be interpreted as anxiety over the rates paid by machine translation agencies. I have repeatedly stated that the main argument against post-editing is not Luddism but rather low absolute wages. I would like to hear machine translation “apostles” speak more loudly in favor of high wages for post-editors. The crushing silence in the regard leads people like me to suspect that the pro-MT movement is really a slightly more techie version of cheapo translation. And of that we have enough, can’t we agree on that?

  2. Fabio Salsi

    I quote: “No other freelance profession (journalist, electrician, car mechanic, plumber, even lawyers!) would resist a tool that enhances productivity/turnover…. even if it meant lower rates but more work…”

    Really? If a tool enhances productivity but the gain does not translate in more money per hour for the people who have to use it, what is the point? Why would I be interested in it as a (freelance) translator? If a translator chooses the path lower rates but more work is a downward spiral. As company you might follow that path but as a freelance translator your time is limited, you cannot compensate for lower rates by working more. I would rather choose a voice recognition technology which allows me to double my speed and double my income rather than MT where I would actally lower my per hour rate. As they are MT pricing schemes are an economical suicide for a freelancer.

    1. pangeanic

      Let me reply to both arguments (which I really appreciate) in one. Unlike many MT apostles or “evangelists” I am not a computational linguist. My training comes from mechanical engineering and languages. I began my career in translation when the best we could do with terminology was to use F4 in WordPerfect to copy/paste certain expressions that needed to be consistent in the document. For many translators, it was a golden era.

      I introduced the point about machine translation in a wider discussion – literary translators (where translation is really a “work of art”) have been historically the lowest paid, working in-house for many publishing houses or freelancing. They are not paid by the word. Only professional translators are. The fees that some of us use to be paid as freelancers have become agency fees nowadays – the trend is downwards. Refer to studies from Common Sense Advisory, for example.

      Perhaps because of my training, I approach translation problems in a pragmatic way: if machine translation lets me do the work faster (even at a cheaper rate), I will take it. True, there have been too many abuses, even from the times of Systran 4 and 5 where MT has been misused to lower prices. I argued about the “cost down” approach in this blog some months ago. Machine translation is a tool to help get the job done (again, refer to Most translators have not a problem using GT or Bing to help them, they have a problem with post-editing. On the other hand, technology developers (I had a particular interest in the technology and thus started building machine translation tools some yars ago) look at things from the inside, they are aware of the effort required to get the sentences joined. Personally, I’m done with translation – whenever I have to lend a helping hand to anybody at Pangeanic, I ask for a pretranslation so I can concentrate on
      a) correcting the obvious machine (calculation or rule) errors
      b) not having to type
      c) not having to “think” word translations which have been thought and worked out for centuries – I prefer to use my grey matter in the embellishment of the final text and spend more time on QC.

      This frees the translator to do other things. Simple math will prove that 1000 manually translated words by a professional typing translator at 6-7-8 cents equal 5000 of a good MT output at a reduced fee. A fair payment scheme/guideline is argued for in and in the whole FAQ section deals with many doubts and fears.

      I see all future in machine translation and in professional translators adopting a dual role as post-editors in controlled environments and full translators when the engines are not performing well (although the leaps forward in machine translation technology are greater and greater by the day as more and more data becomes available).

      Regarding the use of a tool that enhances productivity, I always like to give the same example: the walls of my house are painted. They are painted in different colours and each of my children have chosen a different colour. We don’t care how the painter(s) painted the walls. I wasn’t present when it happened. I think they used some sprays, brushes, etc. All I had to do was to choose the colours. I paid a price for the painting job. Then, on those same walls, my wife and I hung some other paintings, photographs, frames…. not of a high value, but they are craft: paintings we bought on a trip to Paris, a large picture of two of our daughters, a nice Chinese calendar. Some of those pictures and frames are worth more than what I paid for having the wall painted – even though the area is much smaller. There is a difference in utility, in area, in value and in use (some walls have become dirty with use and the little ones have painted on them). I think you get the idea of what I am referring to. Machine Translation and Post-Editing is not a trick against translators. It is a different thing to translate a lot of content which sometimes has low value, but has to cover vast areas. Believe you me, translation companies are in the same situation and pressures as freelance translators when it comes to this.

      I would really like translators to be engaged in the discussions joining groups such as Translation Automation in LinkedIn, for example. Some free reading can also be found at

  3. Bob Beer

    Machine translation may be a viable option between related languages like English, Spanish and French, but with a language like Turkish (my second language) it produces complete gibberish in all but the simplest of sentences. I’ve had some texts that were obviously machine-translated sent to me for editing; they were so far off that it was easier to translate them from scratch. Perhaps in technical work it’s more feasible but bit if the language has to flow well and you don’t want to spend your time comparing the original text to the nonsense on the other side.


Leave a Reply

Your email address will not be published. Required fields are marked *