Machine Translation: The Polyglot Brainchild

Jala Translate
Sep 26, 2018 · 5 min read
Image for post
Image for post
“white robot illustration” by 수안 최 on Unsplash

As the world grows more digital and more interconnected, so grows the demand for translation. Thanks to the internet, information has never been more accessible, yet language is still a barrier for much of the world’s non-English speaking audience. Across regions, across countries, and across continents, the need to understand and be understood has become an ever constant necessity.

Despite the volume of content requiring translation, traditional translation methods are still very time-consuming. The speed of human translation varies widely, and depends not only on the translators, but also largely on the language pairing, the genre, and complexity of the text. For professional translators relegated to working on paper alone, translating just 2,000 words per day (assuming a day consists of 8 hours of work), is a respectable rate. But that’s an entire day’s worth of work spent translating less than the length of an average undergraduate essay.

Image for post
Image for post
Photo by Chivalry Creative on Unsplash

Translating with the help of tools and resources such as Computer-aided Translation (CAT) software and Translation Memory, can help to improve a translator’s productivity as well as their word-per-day output. Having access to a database of translation terms, particularly specialised terms tailored to the genre or domain of the text makes life much easier for today’s translators. Nevertheless, for a translator to interpret a text, word by word, takes time and patience. This process is not only slow, but also potentially expensive.. Professional translation services at Lionbridge (the leading translation agency), range from $0.20/word for a 3 day turnaround, to $0.50/word for a 12 hour deadline. To translate this blog article within 12 hrs would cost more than $300! Of course, the expectation is that the price of translation services comes with a certain assurance of quality. For projects where accuracy and precision cannot be compromised, paying a premium can make sense, but with constant innovation, machine translation is becoming a more feasible alternative every day.

Machine translation is translation that is not just assisted by, but completed by a computer. Humanity has long dreamt of devices that could make communication across languages effortless and instant. The feasibility of this dream was first demonstrated 1954 by the IBM 701, a computer which occupied two entire rooms of IBM’s New York office. It’s worth mentioning that, machine translation was the first non-numerical application of a computer, ever, which speaks volume to the importance humanity places on bridging language divides and facilitating communication across culture. The computer was dubbed by newspapers as a “robot brain” and a “polyglot brainchild”, and with the help of a monolingual operator, the machine successfully translated 49 sentences from Russian to English, prompting the engineers to proclaim that within 5 years, machine translation would become an “accomplished fact.” While the timeframe of the prediction was optimistic at the time, today, effective machine translation has gone from aspiration to reality.

Image for post
Image for post
“white and black personal electronic tansactor” by Anastasia Dulgier on Unsplash

Every day, over 100 billion words are translated on Google Translate alone, instantly and for free. For comparison, you would have to hire at least 40 million translators, pay them $10 billion, and then wait 8 hours to complete the same volume. Compared with traditional human translation, machine translation is fast, it’s free, and, with new innovations in machine learning, it’s improving year by year. In September 2016, Google unveiled the Neural Machine Translation system, which replaced its previous Statistical Machine Translation System. This new system translates by entire sentences rather than word by word. Using the context provided by the sentence, it determines the optimal translation, and then reconstructs the translation to adhere to the grammar an actual human would use. This is made possible by artificial intelligence and techniques such as Deep Learning. The results have been almost unbelievable. Overnight, it had improved to a degree almost matching the improvement the old system had made over the course of its 10-year life span. At the inauguration of the new Google office in London, Google CEO Sundai Pichar presented an especially illustrative example of the leaps and bounds the new Neural Machine Translation system had made over the old system: The quote from Argentine author Jorge Luis Borges “Uno no es lo que es por lo que escribe, sino por lo que ha leído” was translated by the old Google Translate system as: “One is not what is for what he writes, but for what he has read.” The meaning is not at all clear, and the way the sentence is constructed makes it quite obvious there was no human involved in its conception. By contrast, the A.I. powered translation produced a much clearer, and natural sounding result: “You are not what you write, but what you have read.” Of course, this example was selected because of the massive improvement on the old system it demonstrates, but the point is clear, Google has made an enormous leap. Today’s Google Translate, which runs on any device with internet access, rather than a computer that needs to occupy two rooms just to operate, may finally deserve the names “robot brain” and “polyglot brainchild.”

Yet even with all the cool technology involved in the latest edition of Google Translate, it is still prone to make mistakes no human translator would ever make, especially when the text in question gets longer and more complex.

At Jala, we recognise that massive strides have been made in the field of machine translation, but we also know that to ensure quality translations are produced, a human still needs to be involved somewhere along the way. Jala uses machine translation to produce an initial translated result and provides access to other useful tools like translation memory, and text segmentation, which can help translators enhance their productivity dramatically. For complex translations where this might not be sufficient, users can connect with Jala’s community of polyglots, who can offer their language ability and translation expertise to ensure that projects are properly translated. By combining the power of modern machine translation, with the knowledge of real human translators, overcoming language barriers is no longer a distant pipe-dream.

By Mike Chen

Medium is an open platform where 170 million readers come to find insightful and dynamic thinking. Here, expert and undiscovered voices alike dive into the heart of any topic and bring new ideas to the surface. Learn more

Follow the writers, publications, and topics that matter to you, and you’ll see them on your homepage and in your inbox. Explore

If you have a story to tell, knowledge to share, or a perspective to offer — welcome home. It’s easy and free to post your thinking on any topic. Write on Medium

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store