Thursday, 15 April 2021 12:00

A Starter’s Guide to Machine Translation Post-Editing: What to consider when considering post-editing?

Written by Jasper Pauwels

alex knight white robot unsplash

Few subjects are as controversial among translators* as machine translation post-editing (also known as MTPE or simply post-editing). Many translators say they wouldn’t touch machine translation with a bargepole, while others argue it is the future of translation and resistance is futile. In this blog post, I do not intend to advocate any particular position. I do recognise, however, that machine translation is here to stay and that the technology will only become more sophisticated and more widely used. I think that language professionals should be able to make an informed decision about post-editing, and I hope this blog post will help you make that call.

To this end, this article examines some of the drawbacks and advantages of post-editing from a freelance translator’s perspective. Furthermore, it provides a few practical tips on how to approach offers of post-editing work for those considering giving it a go. What post-editing means or how machine translation works is outside the scope of this blog post.

Drawbacks

One of the most striking issues with post-editing is the lack of work satisfaction. I have yet to meet anyone who enjoys post-editing or takes pride in working with this technology. Post-editing machine translation is often compared to fixing a very poor translation. Machines make entirely different kinds of mistakes and do so more repeatedly than any human translator ever would. For this reason, post-editing involves correcting the word order or inexplicably inconsistent terminology, among other unexciting things, and it quickly becomes tedious.

In addition, the purpose of machine translation is to make translation cheaper and quicker. As often cited, you can only pick two out of three in terms of cheap, quick and good. The result of your post-editing, therefore, is expected to be ‘fit for purpose’. A post-editor rarely has the time, nor are they paid, to go beyond fixing grammatical issues and the worst terminology problems. Improving the readability of the text, adapting the message to the readership or just making the translation prettier is more often than not out of the question. Yet these are precisely the areas in which a human translator can excel and get work satisfaction.

Strangely, the fact that it is unpleasant work does not mean post-editors are paid more. Quite the contrary, rates are often slashed, with the argument that the text is already ‘pre-translated’ and will take less effort to complete. This, in turn, means less time is available for the translation in order to earn the same amount. The focus shifts from translation quality to time spent, which again is very unsatisfactory.

It should also be noted that although there are similarities, post-editing requires a different skill set to translating or editing a translation because of the nature of machine translation. Post-editing doesn’t necessarily make you a better or more experienced translator. Translation is a craft best mastered through translating and both new and seasoned translators should be aware of the risk of overly relying on a machine. Machine translations can and will produce unusable output and you need to remain capable of translating yourself.

Advantages

The above drawbacks do not make post-editing very alluring and the advantages for the translator are often overlooked. Of course, the pay for post-editing is often much lower compared to human translation. But if you find yourself experiencing lulls in your work schedule, it can be interesting as a new line of additional revenue. To put it bluntly, poorly paid, uninteresting work is sometimes better than no work at all.

Furthermore, as I mentioned before, the quality of machine translation output is expected to improve over time, which should mean you will need less and less time to make the result acceptable. This higher productivity will result in a better pay per hour, which is fair, considering the results of your efforts will most likely have been used to train the machine and help it to learn from its mistakes. Indirectly, you will have contributed to that productivity increase.

Although striving for less than perfect quality can be very frustrating, you can use the lower expectations to your advantage. Since this doesn’t need to be your finest work, you can put less time and effort into it, saving you time and energy to spend on other, more challenging tasks. I am a morning person myself and I prefer to work on revisions and actual translation in the morning and postpone post-editing to the late afternoon when my concentration starts to drop. Needless to say, you should always meet the agreed quality standards, but post-editing doesn’t require the most productive moments of your day.

Lower quality standards can also mean more room for error. If you have no particular specialisation yet, post-editing can be a relatively safe way to dip your toe into the water. You can try out something new as a post-editing job, whereas you would feel daunted taking it on as a translation job. It may not help you become a better translator per se, but doing your terminology research to check whether the machine translation output actually makes sense can be a valuable learning experience.

Practical advice

Suppose you have considered the pros and cons of post-editing and decide to try it out. The first thing you should do is set clear ground rules with your client. If the client is asking you to accept a lower rate for post-editing, it is only fair to point out they cannot expect your usual quality.

Furthermore, machine translation handles some types of content better than others and it is important to remember the technology’s limitations. Post-editing simple, repetitive technical texts, like microwave manuals or help instructions, may well be twice as fast as human translation. Post-editing highly visible marketing copy, containing puns and cultural references, is probably a recipe for disaster. Discuss your concerns upfront and agree to be paid your usual rate if the machine translation output is indeed disappointing and you have to start from scratch.

Make sure you have a healthy mix of post-editing, translation, editing and maybe other work. As you almost certainly won't enjoy post-editing, you should make sure you keep working on interesting projects as well. In addition, keep in mind that you are not translating the text and you are aiming for a lower quality bar than usual. Nobody will complain about quality being too high, but your efforts won’t be reflected in your invoice.

If you don’t know what to expect from a human translation project, you would usually ask to see the source text. Similarly, it is perfectly normal to ask for a sample of the machine translation to assess if the output is any good and whether the suggested rate is appropriate for the effort needed. Finally, asking for feedback on your post-editing is just as professional as asking for feedback on your translation work.

Closing remarks

I would like to point out that I strongly believe the technology will not put translators out of work. As has been said by others before, machine translation is only a threat if you translate like a machine. Human translators will remain the best option for attractive marketing copy or highly specialised translation.

Moreover, it is important to note that machine translation will only render useful results after being fed millions and millions of words’ worth of high-quality human translation about that particular subject. In other words, machine translation needs human translation. If a client launches a brand-new, revolutionary product, machine translation will be of hardly any use because of the lack of previous human translations. Although the AI experts feel confident that machine translation will reach near-human quality in the years ahead, they do recognise the final 5% required to reach actual human translation quality will be extremely difficult to achieve. I dare to wonder whether human translators will ever be taken out of the equation completely.

In this article, I have taken a very pragmatic view of post-editing, but I am by no means a ruthless businessperson looking for a quick buck. On the contrary, I am a professional translator who takes pride in his work. I prefer translating to post-editing any day of the week. That said, I do think freelance translators should consider machine translation post-editing as a business opportunity rather than a threat to the profession. However, if you feel you would be deeply unhappy post-editing and have enough work coming in, or if it just doesn’t work for your specialism, the disadvantages will probably outweigh the advantages. If you could use a little extra work, it may be worth considering it in earnest. Finally, remember that you are a freelancer and the choice is entirely yours.

*When I use the word ‘translator’ I mean both translators and editors, considering the merits of post-editing for their work.

Read 877 times Last modified on Thursday, 07 October 2021 21:08

Previous articles