Post editing of machine translation will become the defacto standard before long. There, I’ve come to terms with it and I’ve said it, but can everyone else?
As the rise of machine generated translation keeps increasing at its current pace, the translation community must recognise that this may very well become the normal starting point for human linguistic involvement.
Whilst some translators adopt the “head in the sand” approach – and we know who you are – others are beginning to embrace the technology and get ahead of the game, and everyone involved is cutting costs in the process.
Most localisation companies will have to create a Post Edit Machine Translation option, while the linguists supporting them will also have to support this proliferation of technology. The time for filibustering has to stop. Customers are demanding it – and as the old adage goes, the customer is always right.
I am not saying human translation will become redundant (far from it), but it will find its place in the right solutions – one of which will be something you cannot avoid, PEMT.
Are you on-board the PEMT bandwagon?
I agree that this trend is something that will only gain further momentum. I also think that LSPs that learn to work with MT effectively will develop competitive advantage and become more attractive to customers.
The best systems will also be those that are “steered” or guided by savvy language professionals and linguists. As a long-term developer of SMT systems I can assure that the best systems will always be those that have a human touch and that this technology should not be thought of as a replacement for humans except where quality does not matter.
I have blogged about this in some detail at http://kv-emptypages.blogspot.com
We should come up with a good term for this. I think PEMT is not going to be popular. Here are some other possibilities:
MTE = MT Editing
MTC = MT Cleanup
or MT Post Editing without the abbreaviation (via @jerombot)
.-= Kirti Vashee´s last blog ..Linking Translation Quality to Business Purpose =-.
hey wonderful article, I will use it to try to close a new deal. Thanks! But no tools to share your wonderful story. I think your traffic would increase if you add just these simple functionalities to your site. Let me know if Paginar USA can help.
Warmest,
Pablo
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?
This is an interesting question which does not result in a straight forward answer.
The obvious downside, especially for businesses, is that machine translation is literal rather than stylistic so seems as though it has no sense of logic. It is therefore relatively easy for word for word translations to cause embarrassment or even offense. This is why an element of human involvement is critical if you want to get your message across accurately in multiple languages without it having a negative effect on your brand.
However, at the opposite end of the spectrum machine translation can only improve with continued use. For every translation that is run through machine translation engine the more the system ‘learns’. This will inevitably lead to the quality of machine translation becoming increasingly accurate (though it will never be 100%). However, if businesses wish to continue to project a professional image, the need for human involvement will never truly be eliminated. There is no accounting for the detrimental effects that could follow an inaccurate translation and a bit of human intervention can make a huge difference!
In conclusion, machine translation will ultimately drive down the cost of translation over time. But elements such as style, tone and personality won’t be conveyed and there will always be margin for error. Depending on how important accuracy is within each document will weigh heavily on the involvement of machine translation. It might be ideal for user generated content, translating customer emails, etc but for something like a press release, where your key messages are crucial, you need an experienced human translator. And just to add another option to the mix, you can use post editing of machine translation where a editor simply cleans up the text so it makes sense, but it will still lack the tone and style of the original.
For more information about a blended approach that includes machine translation, editing and human translation please visit http://www.appliedlanguage.com/translation_services/blended-approach-.aspx
I’ve been hearing that for years – MANY years. The first time I was told about this issue was when the company I was working for started evaluating the use of machine translation using the Systran system… on a mainframe! It must have been 1983 or 1984, I believe. I showed them how stupid it was, because even typing it on a typewriter (and even writing it by hand) I managed to exceed the speed of this machine translation, and providing a much higher quality.
There is a market for machine translation (MT), but I do not fear that it will displace me. Having a degree in computer sciences I know exactly the difficulty of programming this kind of system, and as of today -and probably for decades to come- machine translation will not be a major issue.
I’ve tested dozens of automatic translation systems & I am not exactly thrilled. The state-of-the-art is not yet “good enough”. (Note: For those interested in free MT tools, there is a nice list in http://www.freelance-translator.info/machinetranslation.php)
Post-editing might be an answer? For certain texts, yes. But not for the type of jobs I usually do. I am pretty confident that I will retire over 10 years or so from translation with the debate still going on… as it has been for the past 26 years since I heard it for the first time!