AI, MT, neural translations – how far away are we today?

machine translation, AI translation, neural translation

For several years we’ve been promised that artificial intelligence (AI), machine translation and neural network translation systems will transform the way we translate content. But how close are we to handing over our translation projects to the robots?

Here we explore where we are now and what the future holds.

Do you still need a human translator?

I’ve seen a few headlines in the last couple of years reporting that AI-powered translation technology is ready to supersede human translators. For example, ‘Translatotron’ Google’s prototype AI translator is reported to translate the tone of voice as well as words. This speech-to-speech translation system tries to capture the infliction in the speaker’s voice, as well as the words themselves. While it’s not perfect, AI technology ‘learns’ from its users so, in time, this could become a very useful tool for communicating with people face-to-face.

Text translation has also come on by leaps and bounds. Neural network translation systems are improving accuracy significantly compared to phrase-based machine translation. Neural machine translation (NMT) works using two different neural networks where one tries to understand what the sentence means, and the other finds the appropriate text in the target language. This technology has the potential to eliminate those mistranslations when a word has multiple meanings as, in theory, the neural network will understand the meaning of the word based on the overall sentence. Google Translate now uses this technology and claims to have reduced errors by 80%.

Machine translation tools like those mentioned above, and the technology we use at Comtec, can reduce translation times and costs, and are very convenient. If you’ve got large volumes of content to translate, or a quick turnaround is needed, machine translation will save you time and money. However, we would always recommend that a human translator proofreads and edits the translation.

AI translation is also beneficial for translating communications from a different language into English, for example, if you receive a message or email from an overseas customer. While the result may not be 100% accurate, you’ll get the gist of what the sender is saying and can then decide whether you need to ask a bilingual colleague or translator to draft a response.

AI can’t translate context, emotions and culture

However, localisation still remains the preserve of the human translator. It’s all about context, emotions and culture. While Google’s Translatotron may imitate the speaker’s tone of voice, this is still open to mistranslation. For example, in China, it’s considered essential in a business setting to show hospitality. It’s not uncommon for a Chinese host to urge you to “Eat, eat, eat” during a business lunch. This style could sound like an order if translated using a speech-to-speech AI tool, which in turn could cause confusion and even offence. However, a professional interpreter would understand Chinese culture and business etiquette, so would interpret it sensitively and put the interpretation in context for their client.

Similarly, machine translation tools will not tell you if the source text is culturally appropriate or not. Or whether a marketing message using emotive language has the same impact in your target market as it does in the UK. For these kinds of translation projects, you need more than translation, you also need to localise the content for the target market. AI translation tools haven’t learned to do this yet, but perhaps in the future…

So when you work with a translation partner, you’re benefiting from their cultural knowledge and awareness as much as their language skills. You can, therefore, expect the final translation not only to be accurate but also localised for your target markets, making it that much more effective.

Get the benefits of AI and human translation

I don’t think it’s a question of AI vs human translation. For some projects where context, emotions and culture are not relevant, for example, when translating a technical manual, there are definite advantages to speeding up the process with machine translation. Semi-automated translation where the bulk of the text is machine translated and a human proof-reader checks it over, is faster and more cost-effective than human translation by itself.

However, for high-impact content such as creative and marketing materials, training programmes and business communications, localisation is an essential factor that only a human translator is currently able to do.

To find out more about the technology we use at Comtec to reduce costs, turnaround times and your stress levels, click here.

If you would like to discuss our translation technology in more detail, please get in touch.

New to outsourcing your translation requirements? Download our free guide below that explores how to get value from translation service providers.

translation guide

Comments

comments