Machine Translation – Different Quality Levels for Different Scenarios
What is machine translation? How good is machine translation? How does it compare to human translation? Do different machine translation engines produce different quality? How secure is machine translation technology? Where do human translators fit into the picture? In this blog, we answer these questions and more.
Machine translation – technology on the rise
Machine translation is the name given to the process of using software to automatically translate text or speech from one language to another. When used correctly, it can be an excellent tool for increasing the speed and efficiency of translation projects, making significant cost savings on translation spend.
Machine translation is growing in popularity: online searches for ‘machine translation’ have increased 16 per cent in the last five years, while the search term ‘machine learning’ has grown 135 per cent. Here at The Translation People, we’ve experienced a more than 2000% increase in the volume of work we carry out with machine translation in the last five years, while website enquiries for this service have increased over 1000%. Those in the know predict that by 2025, 10 per cent of global enterprises will use an enterprise translation hub-type architecture, with 75 per cent of the work of all translators expected to change from performing translations to reviewing and editing machine translated output.
Different standards of machine translation engines
Not all machine translation engines are of the same standard, and some freely available machine translation products can produce questionable results. Not only do we occasionally see funny or badly translated signs and menus, but some of the biggest companies in the world have been embarrassed by machine translation errors: Amazon’s recent Swedish launch, for example, made headlines for the wrong reasons when the translations outputted were lewd and vulgar in nature.
We can therefore see that, whilst free Machine Translation software has obvious benefits, the quality produced can be limited and may need serious reworking by a human translator before it is of an acceptable standard for business use, especially with regard to services like website translation.
Moreover, one of the main issues with this kind of tool relates to data security: with many free Machine Translation engines, data that is loaded into the tool is then recycled within the software to improve its output and could be made available to others in the future. These tools are therefore not suitable for confidential data.
Thankfully, more secure Machine Translation platforms are available, such as the ones we use here at The Translation People, which ensure that data is not shared with third parties. They can also produce better results than some freely available machine translation engines.
Enterprise-level machine translation
We also work with enterprise-level machine translation engines, which can even be trained with client-specific data (previous translations and terminology glossaries) in order to produce even better results that are adapted to a company’s specific language and terminology. Although this ‘training’ does require significant upfront investment, this can pay dividends in terms of higher levels of accuracy and reduced post-editing required. The technology also allows, in some cases, for the engines to be re-trained with human edits to the translations, improving the output even more. For large-volume, ongoing translation requirements, the investment cost of training a machine translation engine may be outweighed by the reduced post-editing costs, especially over longer periods of time.
Machines can only get you so far (for the time being anyway), so whilst the quality of some machine translation engines is of a higher standard than others, we always recommend that output is reviewed by a human post-editor to ensure that the translation is ready to be used in a business context. Post-editors are specially trained to be able to identify improvements to machine translations to achieve the desired levels of quality.
A hybrid approach – the best of both worlds
Businesses should work with a translation agency that collaborates to understand the scenarios and the types of content where machine translation would be beneficial, and where a more human-based approach to translation is more appropriate, which usually depends on the type of content and language combination required. Services such as transcreation are much better suited to a human-led process, whilst technical translation is particularly well aligned with this technology. In many cases, a hybrid approach combining machine translation and human post-editing can deliver high-quality translations at a competitive price.
Companies that invest significantly in translation each year should consider their long-term strategy; working with a provider that combines an enterprise-level machine with humans that are central to the process will achieve time and financial efficiencies, as well as a premium level of quality which achieves all objectives – the perfect combination.
You may also be interested in