Is Machine Translation any good?

In the beginning

The field of automatic translation goes back further than you may think. The first research programmes exploring machine translation systems were carried out in the early 1950s and focused on the Russian<>English language pair. As you can imagine, progress was slow and the quality of the output fell well short of that produced by a human translator. Although interest in machine translation remained constant and a variety of different methods were tried by countries the world over, from a commercial standpoint, translating and editing the ‘old-fashioned’ way would could continue well into the 1990s.

At this point it’s important to make a distinction between machine translation (MT) and computer-assisted translation (CAT). Whereas a machine translation service aims to automatically generate a target language text without the need for any human input, computer-assisted translation software tools are designed to support the professional translator in their work, by storing previously-translated content in a translation memory (TM), integrating glossaries and termbases and applying quality control tools such as spell-checks in the interface directly. By the mid-1990s, CAT technology had made major advances and its use on personal computers started to become more widespread. Today there are a wide range of CAT programmes available and any good translation company will know how to maximize these tools to improve the quality, consistency, and cost-efficiency of their services.

Current practices

Machine translation companies’ ultimate aim is to bypass the human translator altogether. Among the most well-known operators in this field is Google translate, whose online service processes over 100 billion words per day. Impressive numbers indeed. However, you don’t need to be a professional translator to see that the target language text generated still leaves a lot to be desired. Nevertheless, whereas machine translation output from 10 years ago would range from the hilarious to the downright bizarre, today’s tools produce much more ‘readable’ output and can provide a genuinely useful service if your goal is to get a basic grasp of foreign language text or to make yourself understood in a foreign tongue.

However, the problem remains with using machine translation to generate high-quality language that can be used to sell your business in foreign markets. Put simply, despite the progress made, MT programmes aren’t yet up to the task. Nevertheless, the proliferation of services available and the significant improvements in quality have given rise to a new service – machine translation post editing. This involves a human translator performing an extensive editing of machine translated material to make it acceptable for publication. As an increasing number of localization companies offer it as a service, its use is becoming more widespread and is set to grow significantly over the coming years, particularly for more ‘formulaic’ and repetitive text types. That said, if your document requires any element of creativity, such as in marketing and publicity materials, a human touch is always best.

Future perspectives

One thing’s for sure: the machine translation boom shows no sign of slowing down. As the tech giants battle it out with one another for market supremacy, the service available to the end-user gets better all the time. Although for the time being the best results always come from a human translator, either through the traditional method or via machine translation post editing, current estimates are that machine translation will have ‘caught up’ with human within the next 10-20 years. Watch this space…