Machines and humans working in harmony – is this the future of translation?

Experiments in machine translation have focused, understandably, on teaching computers to translate. In recent years, they have shifted to enabling the machines to teach themselves to translate, with AI and deep learning resulting in some good progress. We now have a wide array of translation apps, from sites where you can drop in a weblink to enjoy a full website translation to mobile camera apps that will read and translate a typed or hand-written sign. 

The problem(s) with machine translation 

Despite all these technological advances, machine translation remains something of a poor relation to human translation services. It’s not just that machine translation results in clumsy grammar and some bizarre phrasing. The process is also unable to apply the level of thought that humans can when it comes to shaping a translation. 

Let’s look at marketing translation as an example. Marketing techniques vary from country to country, based on the culture of each audience and what drives them to engage with a service or buy a product. An experienced marketing translator will have a wealth of knowledge that they can apply to the process, subtly shaping the translation to stay true to the original while also appealing to the new audience. They will also be able to suggest elements of the copy that should be localized or even transcreated in order to be appropriate in the target language. A machine can do none of this. 

A new way forward

However, that doesn’t mean that the huge potential that machine translation holds will only ever result in its use by hungry travellers who need to translate restaurant menus. In fact, according to a recent article from Techcrunch, the key to unlocking the real power of machine translation could be combining it with human talent. 

The article cites the work of a company called Lilt as an example. By blending machine translation with human translators, Lilt claims that its translators can translate five times more words per hour than the average human translator. Essentially, the model is like sentence by sentence version of post-editing machine translation

Post-editing machine translation is another example of machines and humans working together in order to deliver a more efficient translation process while not compromising on quality. Tomedes believes that such services will form a core part of the translation industry over the coming years. 

The bottom line

Ultimately, clients who want their documents translated appreciate efficiency and value for money. Services that can translate faster yet maintain the same high levels of accuracy that experienced human translators can deliver, will always do well. They cost the client less (due to the reduction in man hours) and provide results in a shorter timeframe. Key to this is that the quality of the translation does not suffer, as it does when translations are undertaken solely by machines. 

New translation skills

For translators looking to stay one step ahead, then, delivering post-editing machine translation services is a logical next step in their careers. It’s a different process than working with only a source text, as post-editing machine translation means that the translator has both the source and a poorly translated version of it from the outset. However, once that quick has been accounted for, the process ultimately comes down to the same thing for the translator – working to produce a flawless version of the original text in whichever language the client needs.  

Final thoughts

Are you already providing post-editing machine translation services? If not, do you plan to begin doing so in the near future? Or will there always be sufficient demand for the more traditional style of human translation? Leave a comment below to share your views!