Home / Translation practice  / Translating 1,000 words per hour: is it possible?
  • Paul Filkin

    It’s unfortunate that the comparison was based on Studio 2011, already no longer supported, and the translators didn’t make use of any of the features that could be used to improve productivity (based on comments in the report). The results are of course still impressive and do support the arguments for using machine translation in one way or the other, but I wonder how these would compare with any modern tool where the translators knew how to make the most of the productivity features in the tools being used (Trados, Studio, memoQ, DVX or whatever), and where the project managers prepared files beforehand with similar knowledge of what can be gained.
    I definitely think the technology is great, and it’s a natural extension of what so many companies are doing already, but I don’t think the comparison is a realistic one, and certainly shouldn’t be used as the basis for a new baseline of what translators can be expected to achieve.

    26 April, 2016 at 13.04 Reply
    • Kevin Hendzel

      These comparisons of translation speed improvements based on MT/CAT hybrids remind me of arguments over improving air travel by comparing different balloon sizes and gases in the early 20th century.

      Uh, guys, forget the balloons. See those two brothers with their flying machine?

      Translators working in fields they know exceedingly well — most often technical translators — have been producing at 1,000 words per hour for at least the last 3 decades.

      It’s called dictation.

      With the advent of accessible and affordable speech-recognition technology in English and the more common European languages, that feat now only requires the translator, the software (Dragon Naturally Speaking generally outperforms rivals) and his or her microphone/headset.

      The beauty of this technology is that it is BLISTERINGLY fast.

      As an experiment, just to see if I could outrun the software, I once dictated from finished English text (so, not a translation) and was able to vocalize 179 words per minute, with only 3 errors in the final software-produced text.

      179 words per minute = 10,740 words per hour. Again, not a translation, but this software is brutally and elegantly fast.

      The future is not MT+CAT, it’s much more likely to be voice-recognition technology that relies on prompts from previous translations and finished text.

      It is possible to get speed and quality at the same time.

      That product is called wetware. 🙂

      More details here (P.S. The translator keeps all cost benefits and leveraging from dictation in his or her own pocket).


      26 April, 2016 at 23.40 Reply
      • Paul Lambert

        I’m with you on that one, Kevin.
        Now that my volumes are greater than ever, I have earnestly taken to using Dragon to dictate as much as possible. The biggest obstacle in the begginning is the inaccuracy, but the software learns to recognise my voice over time and I spend less and less time correcting words that the computer misheard. As I need to review and edit the text anyway, I find that dictation plus a careful proofreading still saves me more time in the final analysis than would typing plus proofreading.

        Translating Swedish to English, I can dictate about 2400 words per hour (raw terms – before editing). Danish and Norwegian take me longer as I am not as fluent, but even then I easily come in at 1500 words per minute.

        Given that I am at a stage now where I turn away more work than I accept due simply to volume, I see dictation as the only means forward in my business.

        25 September, 2017 at 11.25 Reply
  • Nigel Wheatley

    From the Fifth (2009) edition of “A Practical Guide for Translators” by Geoffrey Samuelsson-Brown: “An experienced translator is able to dictate around 2000 words per hour […] but being able to dictate [sic] more than 6000 words a day, regularly, is extremely demanding.” […] “When I started working as a freelance translator, I wrote out my translations by hand. They were then given to a copy typist who gave me a draft to edit. […] I suppose the effective rate was 200 words per hour. I then progressed to dictating tapes, which were transcribed by an audio typist. Our combined effective rate went up to around 1000 words per hour, including proofreading and editing.”

    26 April, 2016 at 23.57 Reply
  • Martha Hobart

    It’s unfortunate that it’s not possible to test it without creating an account. I’m wary of websites that force me to give them my information just to find out what’s in there.

    30 April, 2016 at 10.15 Reply
  • Birgit Bonde Jensen

    In the conclusions to the experiment nothing in mentioned about individual human factors that would make any projections of the results merely a guess. The fact that you can translate 1,000 words in one hour does not mean you can do the same for the next seven hours. Brain fatigue and high stress level will either slow you or your creativity down. I might do 7-8,000 words in one day, but definitely not the day after.
    Despite autosuggestion much depends on your ability to type, how you use the keyboard and your type of keabord – unless you use speech recognition software (not available in my native language). Or it simply just depends on how your mood is.
    Also, very few freelance translators not working as in-house translators can dedicate the whole day to translation. In a hypothetical 8 hours working day freelancers also reply to mails (and they can be quite a lot), communicate with clients, make offers, issue invoices etc. Not to mention the source text quality. Ever tried a 50 pages manual written in (bad) English by Japanese?
    I know that the answer is that Lilt – or any other software for that matter – boosts your productivity. To that extent, I agree, and the parameters in itself for this experiment might even be reasonable, but the set up environment is unlikely representative. I can’t tell how convincing the experiment is from a sales point of view. To me it looks like just another tool trying to enter a lucrative market and once again feeding the jaws with translators by pretending that translators can be transformed into almost MT-engines and that the translator’s output/throughput is the only lever for increasing productivity on low scale volume.

    1 May, 2016 at 07.46 Reply
  • B Jones

    There are great translations & there are fast translations, but as I constantly have to explain to newbie clients who think we all use Google Translate then send the invoice:
    “My skill set does not include a magic wand.”
    Upshot from this CAT-tool-skeptical dinosaur (full disclosure, I’ll use WF Anywhere, that is ALL):
    Be careful what y’all wish for, because your willfully unseen but ultimate goal is to put yourselves out of work… Brilliant? Hardly.

    3 May, 2016 at 17.12 Reply
  • Kaja Grzegorczyn

    It is of course possible to translate 1000 words an hour, but it is a bit of a worry that discussing it as a given skill of every translator might create unrealistic expectations (in client’s mind, if this is the case, what’s the problem with 8000 words/working day?). We should be careful to stress that other factors influence translation speed as well and no one is able to sustain thousands of words per day without compromising quality (I’ve been there, four days into translating an urgent, long clinical trial document my brain refused to understand English any more…). I agree that it should be up to an individual translator to work at a speed they are comfortable with.

    16 May, 2016 at 10.02 Reply
  • Renate Radziwill-Rall

    If you want to speed up that much, I recommend Google Translate. Besides this joke, I want to point out that a translator is not just a photocopier between original text and translated text. Every word and every sentence has to go the way up to your brain and back into fingers or voice. So there is a limit somewhere. And as long as this is necessary, 1.000 words an hour is an illusion.

    17 May, 2016 at 18.52 Reply
  • Wulf-Dieter Krüger – EUTAC_WDK

    One lesson and so many business simulations as in that once you give company secretaries leeway pinching every penny out of an agreement your company will be doomed like the Japanese company which recently when belly up for having a book MTed without proper post editing.

    However, I believe company secretaries will enjoy reading this article since they have always believed that translations can be done by anybody who speaks to different languages –WHY fo f sake don’t they do th translations THEMSELVES THEN?

    As it provides them with a lot of ammunition for their pennypinching sprees. I might keep repeating myself, however the quality of the source text was not mentioned – eat could be good communication, it could be bad communication because the author the source text did not take his target group into consideration (one of the basic issues in any communication and why communication may fail and if the source text as failed already, how on earth the translator be able to fix that failure at such speeds inferred here without jeopardising his/her health?

    19 May, 2016 at 02.48 Reply
  • Oleg Gordeev

    Typing 1,000 words per hour is possible but it’s raw translation: high-quality editing will take another 1.5-2 hours. I do not feel any considerable shift to a faster turnaround in the translation industry. Agencies tend to send job offers based on 2,000-3,000 words per day. It means their direct clients do not push them too hard either.

    20 January, 2018 at 11.47 Reply
Post a Comment