Machine translation (MT for short), is what you get when you feed a computer a bunch of words and ask it to give them back to you in another language.
And while you can Google machine translation right now and learn a hundred and one things about its modern-day application, our favourite fact about MT is its ancient Arab roots. Believe it or not, Abu Yusuf Ya'qub ibn Ishaq Al-Kindi, an Iraqi, and a 9th-century Arabic cryptographer, is known for developing some of the earliest techniques in systemic language translation. Yet another reason to be proud of our Arabian roots!
Machine translation’s a pretty hot topic right now, so in the interest of riding hot topics, we’d like to drop you a brief education on what machine translation is, how it can save you money, and, (in support of our human translator friends), answer the age-old question of whether or not the machines are coming for your job.
Yalla, let’s learn!
How Does Machine Translation Work?
There are two types of machine translation; rules-based and statistical. In both cases, computer software renders text in one language into another language.
Statistical machine translation is more mathematical in its machination and relies on algorithms to scan and match possible translations from previous work.
Rules-based machine translation relies on a manually built library of dictionaries, grammatical rules and semantic patterns to translate text. The library can be updated to improve translations over time.
And at e-Arabization?
We’re greedy, and we love technology, so we adopt a mix of both methods, depending on the customer, the content, the purpose and need. Here’s how our two dedicated MT services work:
Pre-editing: we receive the text that needs to be translated, and we manually unify all the ‘like’ terminology, neutralising the tone so that when we run it through the MT software, we minimise the margin of error and create a consistent tone.
Post-editing: our most popular service. In this case, we (or our customers) use machine translation to get an automatic, human-free, rough and ready translation of their content. We then hand this version over to our highly skilled human translators, who edit the content to an acceptable standard.
So, you’ve got your project ready, and you’re convinced that machine translation is the solution for you. But we’d be lying to you if we didn’t share the good and the bad, so in the interest of fair and well-rounded education, here’s our take on the pros and cons of machine translation:
The Pros
- Super speedy deadline deliveries
- The low, low price
- Machines can translate text on any topic without studying it
The Cons
- Accuracy issues
- Arabic language limitations
- Specialist fields (legal, political, medical) still need a lot of human work
The pros speak for themselves; speed and cost are two major factors in most business transactions, but let’s dive into the cons and explain the limitations. Machine translation is a game of compromise, and in the case of working with English into Arabic projects, the challenge is steep. While machine-translated European languages perform really well, Arabic is extra tricky because it’s a right to left language with a grammatical structure and character type that’s totally different from any European language.
We’re working on it
Arabian companies are constantly working towards creating better solutions for Arabic, but in light of this challenge, we usually recommend taking a post-editing approach to Arabic language projects. Our in-house Arabic speaking humans are equipped with the knowledge to spot the odd inaccuracies that MT can’t handle, and it’s hard for a machine alone to provide the kind of quality communication that humans naturally deliver.
Another helpful tip
Be selective about the content you submit for machine translation. In our experience, e-commerce projects work well with MT because they usually involve a huge volume of work, frequent needs, and big cost problems. On the other hand, we discourage legal and medical clients from machine translating sensitive documents that require a high degree of accuracy.
Machine Translation in Action
At the beginning of our working relationship with the e-commerce giant, Amazon, we worked off a very modest translation memory of only 2-3% matches across their content. Now, less than a year later, we support Amazon to translate more than two million words per month, and up to 50% of that content is matched from memory.
That’s an incredible 37% increase in machine-translated content!
The final result? Consistent terminology, and thousands of dollars saved each month. Amazon couldn’t be happier. So, if you’re a frequent translation customer, and you need a high volume of content turned around each month, machine translation could be the answer to your content localizing prayers.
And finally...
I’m a Human Translator, Are Robots Coming For My Job?
The short answer? No.
While machine translation has had its fair share of fantastical press, Alexa or Siri aren’t coming to replace you anytime soon. The truth is, machines still need us. Nuances, slang terms, cultural phrases and complex context are lost on machines that still aren’t equipped with or capable of the creativity, complexity and humour that we use to communicate every day.
It’s smart, it’s fast, and it’s learning every single thing we feed it.
Machine translation isn’t a trendy solution that’s going away anytime soon, it’s a developing technology and it’s learning fast. Machines are meant to make our lives easier. Learning to work with them, and using them to enhance and accelerate the work we need to do for our customers is the best way to assimilate MT into our processes and stay ahead of the curve.
Want to know more about machine translation, or want to drop a quick Marhaba our way? Visit e-arabization.com and check out how we use machine translation to turn e-shops into Arabic and connect local businesses with the global marketplace.