Skip to main content

Why we must combine machine and human in the translation world

machine learning
(Image credit: Image Credit: Geralt / Pixabay)

Machine translation can offer significant time and cost-saving efficiencies for international businesses, so it’s no surprise it’s growing in popularity around the world. But for those businesses still to implement this technology for their international communication needs, recent headlines around the inaccuracies of such technology might be a cause for concern. 

For example, a thread of tweets about Google Translate’s capabilities recently went viral and saw them accused of making sexist assumptions about gender-neutral language. There’s also been a study that suggests free translation tools contribute to significant misunderstandings of legal terms with conflicting meanings for some words. Meanwhile, after Amazon recently used its own automatic translation technology to launch its first eCommerce site in Sweden, it hit the headlines when the translations outputted were lewd and vulgar in nature, bringing attention to the launch for all the wrong reasons.

Yet despite some concerns around the accuracy of machine translation, inquiries and uptake in the technology are surging. Online searches for ‘machine translation’ have increased 16 percent in the last five years, while its related search term ‘machine learning’ has increased 135 percent. 

According to a recent report, by 2025 10 percent of global enterprises will use an enterprise translation hub type architecture integrated with IT operations, with 75 percent of the work of all translators expected to shift from creating translations to reviewing and editing machine translated sections. Additionally, over the last six years, UK-based translation provider The Translation People has experienced a 2,016 percent increase in the volume of work it carried out with machine translation, while website inquiries for this service increased 1,130 percent. 

It’s clear that machine translation can offer several benefits to businesses, and the trend of its widespread use is expected to continue rising, but it must be carefully applied and not used in isolation as part of a blanket approach to translations. In fact, the technology only comes into its own – providing the most accurate, effective and creative translations – when delivered as a hybrid solution together with skilled language and translation professionals. 

Standardized approach

Machine translation centers on processing input and producing output. Words, sentence structure, subject and grammatical information are all analyzed, then translated into the desired language. But the accuracy of the output is determined in large part by the size of the language database sitting behind the machine. The more bilingual material fed into the engine by base engineers, the better the result will be, whilst the quality of the output is also determined by the nature of the text, as some content types work better than others. Direct and straightforward language – such as that found in instructions and user manuals – is well suited to machine translation; emotive copy or complex language, less so. 

Currently, there is no universally agreed-upon or standardized approach to machine translation to help ensure accuracy across the board and, until that happens, the technology will always benefit from the expertise of a human translator. Not taking this step can lead to embarrassing errors, such as was seen in East Asia when machine translation was used to create a series of street signs. Instead of achieving complete accuracy, signs stating ‘poisonous and evil rubbish’, ‘pregnant woman over 70 lounge’ and ‘slip and fall carefully’ ended up adorning the streets. 

To achieve the highest quality results, customized engines should be developed for clients on an individual basis. An assigned language expert can train the engines over time, enabling the machine to understand the business’ style of writing, tone, nuances and industry jargon. If any element of a translation is incorrect, it’s spotted and edited by the translator. These edits can be used over time to train the machine not to make the same mistake again in the future, driving a continuous improvement in the output of the machine, achieving a more efficient process and greater accuracy in output. 

Translation providers

Machine translation relies on quality time and expertise being invested to achieve optimal results. Assuming it will work perfectly every time is to risk materials becoming erroneous, and therefore risk your business’ reputation, client relationships and future success. Security must also be considered; using a free, online tool typically involves having to input client data to a cloud storage space where it’s combined with translations from other businesses – all these materials are then utilized to improve the engine’s overall capabilities, so it is unsuitable for businesses translating confidential data or materials. 

To avoid these risks, businesses should work with a translation provider that collaborates to understand the scenarios where machine translation would be a benefit, and where human input is required, depending on the type of text and language combination required. Many texts may not have their objectives met as well with a standard machine translation solution, and clients may save time and money by having an expert, human translator work together with them in a more manual capacity, rather than spending several weeks or months nurturing a machine system. 

In many scenarios, machine translation will offer a sufficient service to translate basic materials to a certain level, and it will continue to become more sophisticated as time goes on. But a true consultative approach, combining technology and human expertise, is required for any material where a client can’t risk anything less than 100 percent accuracy. Companies that use translation every single day and invest significantly into it each year should consider using enterprise-level machine translation in partnership with a provider that places humans central to how a machine translation system develops on behalf of the business. Introducing this as part of a long-term strategy will achieve time and financial efficiencies, as well as a premium level of quality which still puts the human touch central to the translation process.

Alan White, Business Development Director, The Translation People

Alan White heads up the Business Development Team at The Translation People. With 20 years’ experience in the translation industry, his role today centres on advising best practice when it comes to communication with international audiences, by providing companies with customised solutions.