Machine translation is one of the technologies most commonly used by language experts, but it is often misrepresented in the media. Excitement about AI and the prospect of algorithms replacing humans for every task imaginable has created a narrative of machine translation replacing language professionals too.
In reality, machine translation is a tool used by human translators to speed up their workflow so they can deliver maximum quality in a shorter amount of time. Used properly, the technology provides a flexible, fast solution for a wide range of translation tasks, even if it never produces the final approved content.
In this article, we explore the most important best practices for using machine translation effectively.
Like most tools, machine translation is effective at the job it is designed for, provided you use it correctly. Broadly speaking, this means incorporating machine translation into a planned, strategic translation process designed to produce the content you need.
For example, if you are translating news articles from one language to another, machine translation is an excellent tool for quickly producing a first rough draft for human translators to post edit, verify and approve before publishing.
A more complex process could involve translating user manuals for a range of consumer technology products for multiple language markets. In this case, you might use a combination of translation and other related technology (translation memory, terminology management, etc.) to automate the first draft of each language version.
Then, the drafts would be post edited – with light edits or full post-edits – and standardised by human translators to the specifications of the brief, ensuring accuracy and consistency across all versions. And, once the final versions are published, copies are stored for future reference in the case of new product variations, additional languages, etc.
The ideal machine translation process including choosing the right post-editing steps and the role of MT depends a lot on the requirements of the project.
Before using a machine language translator on any project, you need a clear plan for your project and an understanding of how to implement the technology into your translation process.
We recommend following these best practices:
Identify your goals:
Are you using MT to speed up document translation, using it in a live setting for instant translation or for something else?
Consider the input format:
Are you translating a text, live speech excerpts or an audio recording?
Choose the right MT:
Select the right type of machine translation for the job – e.g. text translation, speech translation, etc.
Know the limitations of MT:
For example, complex language is more difficult for algorithms. Background noise in recordings also makes speech recognition difficult.
Optimise the input:
Optimise the input content for the best output results. For example, clear up the audio content to make speech recognition easier.
Use the right tools:
In addition to machine translation, use translation memory and other tools to maximise accuracy, consistency and productivity.
Review process & post editing:
Implement a review process that guarantees 100% quality and accuracy. Post editing can help you get closer to that result.
The limitations of machine translation are more prominent in certain environments than others. For example, the linguistic distance between English and Chinese is far greater than between English and Spanish, making the first conversion more difficult for algorithms.
The input format is also very important. If you are importing text into an MT system and outputting text in another language, the conversion is simpler algorithmically than translating speech in real-time.
In the case of translating live speech, you are relying on algorithmic voice recognition and transcription into text before any translation takes place. And each of these tasks multiplies inaccuracy.
By understanding these limitations and the application of machine translation in your project, you can take proactive steps to mitigate potential issues, whilst making the most of this powerful technology.
The golden rule of machine translation is to never use it as a one-step or standalone translation solution. Even if you are simply using the technology for research purposes, such as sourcing information from foreign news stories or public studies, you have to be aware that the slightest mistranslation could change the entire meaning of the information you source – or the interpretation of it.
In the age of misinformation and disinformation, there are enough issues with inaccuracies and misinterpretations within languages as it is.
More importantly, publishing machine-translated content that hasn’t been post edited and verified by professional translators should be avoided. Companies become responsible for the accuracy of everything they publish, so having the correct translation and editing procedures in place is very important. This will ensure the protection of all target audiences, as well as a brand’s reputation.
In the most harmless scenarios, translation errors may simply lose businesses sales or have them trending on social media for the wrong reasons. However, the consequences, and potential legal ramifications, of translation mistakes in important documents like marketing materials, T&Cs and user instructions can be extreme.
Machine translation is one of the many steps and options in a reliable translation strategy – but never the only one.
If you need help with implementing machine translation into your translation strategy, our language tech consultants can help you. Please fill out the form on our contact page to arrange a meeting with one of them.