Machine translation: good, bad or “ugly”?
![hands on computer keyboard](/sites/default/files/styles/large/public/2021-10/szabo-viktor-5-BnOWM5iQU-unsplash%20copy.jpg?itok=Xt1s_LfJ)
In recent years, a new generation of artificial intelligence-based machine translation has revolutionised the daily lives of ordinary consumers and the work of translation professionals. Automatic translation is available in online services and social media platforms, where it often helps users understand key content. However, automated translations occasionally cause misunderstandings, frustration or even embarrassment, negatively affectingthe customer experience and the organisation’s brand image. The following GPU analysis™ offerssome tips about when and how machine translation could best serve your organisation.
The “great breakthrough” of machine translation has been expected for decades, but until recently trials typically started hopefully and ended in scepticism. Automatic translators were unable to deal with grammar or inflected forms – especially for small Nordic languages – never mind customer-specific style, tone of voice or specialised terminology. Both customers and translation professionals have been concerned about errors in translations and the usability of translated texts – with good reason.
However, the development of artificial intelligence and so-called neural network machine translation (NMT) has significantly improved the quality of machine translations. The big question is how the scheduling and cost benefits of a “translation machine” can be combined with understandable and user-friendly communication and a customer experience that supports the brand.
The good #1: Machine translation is a professional’s best friend
Machine translation is at its best in the hands of professionals. An article published by the research company Gartner estimates that by 2025 as much as 75% of translators’ work will consist of post-editing material that has been machine translated. A quality machine translation makes a translator more efficient, speeds up the translation process, and helps get a product to new markets quickly.
Post-editing of machine translations differs from traditional translation work. In addition to machine translation technology, the most advanced language service companies are continuously developing the actual translation process, new and efficient work methods, and the competence of language experts. This should also be taken into account when selecting a machine translation service.
The good #2: Keep it Simple, Sweetheart
Before introducing machine translation, it’s important to know the purpose for which it will be used. Machine translation provides the greatest benefits when translating unambiguous texts, such as reports, user manuals and instructions. The less interpretation and style requirements in a text, the better a machine can handle it. This makes editing faster and more efficient.
The topic of the text is also an important factor in how useful machine translation can be. Common machine translation solutions are based on language models that are “taught” by feeding the tool enormous amounts of text. This is why machine translation often struggles with unusual words or specialised terms: words that rarely occur in teaching materials make it difficult for the machine to identify their connection to other words. If you’re considering purchasing a machine translation solution, it’s worth ensuring that the technology offers methods for implementing your organisation’s own terminology.
The good #3: Customised machine translation to meet your needs
If your organisation is constantly producing large amounts of content for a specific purpose, machine translation customisation may be a profitable investment. This means that machine translation learns to take the style elements and terminology of those texts into consideration. The requirement for customisation is the availability of large amounts of high-quality and consistent bilingual translation material as well as specification of key terms and their translation equivalents and metadata.
Machine translation is not an automatic answer to all translation challenges, but it’s a useful tool in the hands of professionals. Cooperation between a person and the machine is still needed – before, during and after the translation process. There’s no cause for concern if you need advice, support or people for consultation, preparation and maintenance tasks: an expert language services partner can also help you collect teaching material, process and enrich language data, and handle terminology development work.
The bad: Reputational damage or business risk?
For the time being, machine translation still needs someone to hold its hand: without post-editing, an automatically translated text may contain errors that are humorous or annoying at best – and life-threatening at worst. A bad translation always causes reputational damage, but sometimes it can also be a real business risk.
For some content, cooperation between a machine and a person is just not enough. For example, marketing materials usually require so-called creative translation (transcreation). In this sense, a clumsy machine translation is more likely to confuse a language expert who is trying to adapt style and tone of voice or optimise a multilingual search engine. In such cases, the end result can be inappropriate – in addition to requiring excessive amounts of time and money.
The “ugly”: At times, “adequate” is good enough
Sometimes an adequate machine translation is better than no translation at all. According to research by Gartner, 66% of consumers want to shop online in their own language. An “adequate” translation – for example, an integrated and customised machine translation solution optimised for specific terminology on an online shopping platform – can offer the consumer a good customer experience, reduce the go-to-market time for new products, and increase sales.
An “adequate” machine translation can also be an accessibility and participation factor. In some situations, it may be the only way for a person to get information in their own language. This can be vital in crisis situations. For example, machine translation can help public administration improve communication for all language groups in its own channels, thus ensuring the reliable communication of a message instead of disinformation. The European Union has been promoting the development of a machine translation solution between all European languages for years in different projects. Therefore, machine translation can also represent an important tool for democracy and equality.
Lingsoft is one of the leading providers of language services and language technology solutions in the Nordic countries. It produces a diverse range of translation and language services, with a particular focus on language resources for Finnish, Swedish and other Nordic languages. Lingsoft's machine translation solution and other language technologies are available via the European Language Grid and as an SaaS service maintained by Lingsoft. Contact us to ask more about our services and solutions!