Google Translate is generally quite capable of translating individual words, but the moment you add complex sentence structures or culture-specific expressions to the mix, chances are you’ll be greeted with some rather odd translations.

Over the past 13 years, Google has developed a fair number of complicated techniques to improve the quality of their online machine translation service, Google Translate. Complicated to me, that is, since I’m not particularly familiar with ICT and other such things. Techniques such as neural machine translating, rewriting-based paradigms and on-device processing have certainly contributed to Google Translate’s quality and accuracy. It all sounds so ‘Star Trek-like’ to me, but then again, what do I know?! Still, Google Translate doesn’t generally produce the same quality as a human translator would. That isn’t just our opinion, by the way – many Language Service Providers (LSP’s) would tell you the same. Google Translate is generally quite capable of translating individual words, but the moment you add complex sentence structures or culture-specific expressions to the mix, chances are you’ll be greeted with some rather odd translations.

Machine Translation at Ludejo

Now, at Ludejo, we certainly aren’t opposed to technological developments that’ll expedite and simplify the translation process. Though we don’t use machine translation ourselves (unless the customer has specifically asked to do so, and in that case we always devote extra attention to checking the final translation), we are aware of the fact that translation tools such as Google Translate can be a feasible solution for some translation projects. Consequently, we were happy to hear the news that Google had made several new adjustments to improve their online translation tool. They now use a transformer encoder in combination with a recurrent neural network (RNN). This combination allows Google Translate to translate messy sentences (such as informal language) without slowing down the translation process. Well, that’s the plan at least. And that’s just one of the changes the Google team has implemented.  

If you’re interested in finding out more about the technical wizardry Google Translate uses and how it al works, check out this blog on VentureBeat about “How Google is using emerging AI techniques to improve language translation quality” (I really enjoy searching for online content about everything to do with language, but I’d much rather leave info on ICT and transformative technology to the professionals).

Now I can hear you thinking: “If you know so little about the subject, why write a blog about Google Translation?” For the most part because I want to emphasize the importance and value of technological developments and AI, especially in the translation industry. “What? You just said you don’t use machine translation?” Observant, but this is about something much greater than online translation tools such as Google Translate.

Everyone who knows us knows we are a socially-engaged translation and communication agency (our logo being a heart isn’t a coincidence). Ludejo was built on a foundation of love. We love what we do, we love our customers and we love our team. We believe giving back is important. Amongst other things, we do this by organising fundraisers for charities and non-profits like Translators Without Borders (TWB).

One of the things TWB does is improve communication between humanitarian workers and groups of the population that are in crisis. Especially in crisis situations, it is incredibly important to act quickly; every second counts. Being able to quickly provide a translation is of utmost importance, which brings me back to the technological developments in the field of machine translation: As long as we continue to pursue efforts towards quicker and more accurate translation tools, organisations like TWB will be able to offer help within a shorter amount of time.

In addition, machine translation can help TWB by improving the communication in marginalised languages. In 2019, TWB received financial support from The Cisco Foundation to develop data sets for machine translation for two marginalised languages. The motive for the donation was the successful pilot project for TWB’s Gamayum programme, for which TWB developed a machine translation tool for Levantine Arabic, especially for the World Food Programme. In case that wasn’t impressive and admirable enough, they also managed to complete the project in a mere six weeks (I expect you can understand why the theme of our next fundraiser for TWB will be ‘With Great Power for TWB’, because these people are nothing less than superhero’s).