Is machine translation making a comeback?

By on August 24, 2010

Now, before I get shot down by a flurry of irate translators, hear me out.

There’s been an increase recently in the use of post-edited machine translation for some projects where the volume of content is so huge, and the time window so short that human translation, and then proofreading and subsequent editing of the text, would just not be practical. We at Web-Translations are observing this trend with great interest.

Every week I read an article or comment telling me that that the quality of machine translated text is improving all the time, and having used a couple of the most well-known online tools recently for research and information purposes, I’d have to agree.

However…and this is a very big BUT…the raw product of these tools is never going to be a substitute for professional human translation. There’s just no sensible comparison.

If we’re really in such a hurry to get content translated and published online, using machine translated content and then having it edited afterwards to produce a text of reasonable quality does offer a solution where the client involved is not willing to compromise either a) on cost or b) on speed, or even both of these.

To summarise the benefits of post-edited machine translation:


The cons are (fairly obviously):

The logical next question is, will professional translators accept this development and be willing to edit machine-translated output ready for publication online?

We are regularly contacted by newly graduated linguists and post-graduates who have just completed translation qualifications, eager to gain experience in a sector where most reputable companies require 3-5 years of commercial experience before considering freelance applicants, a tough environment indeed. Would this offer an opportunity for them to cut their teeth and put their skills to practical use? Or is it insulting to trained linguists to ask them to do this, even if they might lack experience?

Another problem rears its head too – how would a layperson know what they were buying? An unscrupulous company could say that they provide professional translation, and actually be making a larger profit by using machine translation and then post-editing (it is rumoured that some companies already do this). There is enough confusion for clients not familiar with buying translation services on what a fair price is and what they should expect from a translation provider, without this added complication.

Where might this type of service be successfully used? It isn’t suitable for translating manuals or instructions, and would not have the desired effect if used for sales or marketing copy, but there are some situations where it could be applied. Here are some of my suggestions – let me know if you agree, or can suggest any more:

News articles

Blog posts


Personal emails

Text messages

Automated messages – e.g. confirming an online order, or acknowledging receipt of a help/support request (these are often translated only once, then automatically sent to the recipient via a database).

Most instances where text will be published would not be suitable scenarios for post-edited machine translation, as there is too much exposure to risk. Imagine if a manual for a chainsaw were translated by machine – even if the text were post-edited by a professional translator, this is something that a manufacturer would not consider, as the potential for being sued would be far too great.

Hopefully this development does not pose a serious threat to translators – if properly used, it could reduce the pressure on all of us for those rush jobs that always crop up, and could lead the way to a more efficient process for certain projects in future.

OK – time for me to don my crash helmet and wait for the backlash of comments…



A statement we often hear from UK-based SMEs is “Why pay for professional translations when I can get it for free on Google and other sites?”. They think they are saving money by using such machine-generated web-based translation services. But are they really doing themselves any favours in the long run?

And there’s more……

Sophie Eadon on Aug 25, 2010 at 2:00 pm

I found lot of things on web for auto translation. Almost all major companies are behind auto translation plugins or widgets to launch. One more recent new regarding translation, google is going to launch a voice to voice translation that means no need for interpreter in future.

website translation on Sep 08, 2010 at 7:09 am

As a company we insist on human translations with human recordings. Having said that, for demonstration sites we use machine translation as the nuances and in fact the meaning is unimportant.

With regard to Google’s Voice to Voice translation, I can assure you that this is a way off having any meaningful business to business application. It is real and I have talked to a member of the Google team in connection with it.

Having worked with leading voice recognition companies, these systems only give effective recognition when they are trained to a given user in a given environment. The systems are language and speaker dependent with English, Spanish, German being the strongest languages and then would offer accuracy of approx 90%. They demand a huge amount of computer processing giving problems of scale. Speaker independent recognition falls way below that quality.

I believe the Google system is initially aimed at phone based communications where fundamental problems with the audio quality further restrict the possibilities for voice recognition.

Add in some computer translation to the mix and pretty soon we are getting something pretty useless other than for the most mundane regularly used phrases in ideal conditions.

Neil McCutcheon on Sep 14, 2010 at 1:14 pm

“The logical next question is, will professional translators accept this development and be willing to edit machine-translated output ready for publication online?”

The answer is clearly no as we don’t even like editing texts mangled by bad human translators. I also don’t agree with your suggestions of use, news articles and blog posts in particular as these require writing skills which even most humans don’t possess, let alone machines.

Olaf Knechten on Oct 27, 2010 at 1:17 am

I’ve noticed that Google Translate gives you an option of suggesting a better translation if you think the machine-generated one is wrong. This, I presume, is used to train the program. I wonder if this training will eventually lead to machine-generated translations which are truly accurate. If enough people contribute, one is tempted to think that the sheer weight of contributions could tip the balance.

Personally, I doubt it; I think that accurate machine translation is either a pipe dream or at least many, many years away. After all, the number of possible sentences is infinite. I suppose the point is that no matter how many dictionaries the machine has access to, it just doesn’t ‘get’ it.

I read a very interesting book by Antonio Damasio a while back called Descartes’ Error and it showed that emotion is an essential part of human reasoning and decision-making and that when the ’emotional’ areas of the brain were damaged (particularly the prefrontal cortex), decision-making became erratic and unreliable. While it’s true that emotion isn’t especially linked to language, I wonder whether the fact that we care about the effect our actions have (we want to carry on living) has created a cognitive structure that a computer, unfeeling and not self-aware, simply cannot compete with. Common sense and language come so easily to us, but they are very hard problems for computer science.

Google are also working on a program to translate poetry. If the attempt fails, which seems likely, the reasons why will be revealing. There’s an article on this in the Guardian:

Greg Hunt on Oct 27, 2010 at 9:45 pm

@Olaf – thanks for your comment, I wanted to see what translators thought, and I do agree with you, up to a point.
Some people feel, however, that online content such as blog posts, comments, social media updates, etc do not to be translated to the same standard as other content such as a website. That’s what I wanted to reflect in this post. Think of a forum/discussion board where there are often errors, typos and slang/text speak are also used – a machine translation can perhaps render this to a high enough level of accuracy for it to be read & understood.
Is there such a thing as “good-enough” translation within that context?

admin on Nov 04, 2010 at 9:46 am

@Greg – thanks for your input.
I also think that a machine won’t ever be able to replace a human translator – surely by the time we have taught it all the rules, they would have changed? The usefulness of this technology continues to develop, however, and I think it’s getting really interesting.

Thanks for the article link, and keep reading!


admin on Nov 04, 2010 at 9:49 am

I think that a significant downside of machine-translated blog posts, comments, or social media updates is that they seriously clutter Internet, e.g. appear in search results and misguide you in various ways. Instead of publishing any machine-translated content directly, I’d rather provide users with a single-click option to have it translated via Google. In this way, anyone who needs translation will easily get it, and those who avoid machine-translated content will benefit as well.


Roman Mironov on Nov 06, 2010 at 9:34 am

I most certenly liked your innovative angle that you have on the subject. I wasnt planning on this at the time I begun browsing for tips. Your ideas were totally easy to get. Happy to find out that there’s an person online that obviously understands precise what its is talking about.

Valentine Bverger on Nov 30, 2010 at 3:36 pm

I do like the way you have framed this specific matter plus it does supply me personally a lot of fodder for consideration. Nonetheless, from what precisely I have observed, I basically hope when the actual feed-back pack on that men and women keep on point and don’t embark on a tirade of the news du jour. Yet, thank you for this excellent point and while I do not really concur with this in totality, I regard the perspective.

Selene Bennin on Apr 16, 2012 at 1:55 pm

Get in touch

How can we help?

By continuing to use this site, you agree to the use of cookies. For more information, read our Privacy Policy

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.