ecommerce-machine-translation

3 myths about automated translation for international ecommerce sellers

An average human translator can translate around 2,000 words per day for 20 cents per word. Not bad, is it? But …

  • … what if your business needs exceed the daily 2,000 word limit?
  • … what if your budget is limited?
  • … what if you need translations in multiple languages?

Wouldn’t machine translation (MT) be a better option?

Of course, even in the case of large translation outputs, you can hire multiple translators, pay them as required and … wait for the job to be finished. And hope that the final translations and terminology used by the various people will be … consistent.

But here’s the deal…

Human translation is not necessarily the best return on investment for your business.

Especially if you sell on international markets and/or you need to change your inventory regularly. Slow translation carried out by humans can cost your business a loss in sales potential.

Human translation has its benefits, but also a serious limit: human capacity. You can employ multiple translators to work on a given project to increase the output, but think again …

… machines can generate thousands of words per minute. And here’s their main advantage: lower cost and consistency.

Even though machines are no linguists, in general they do a pretty good job. Automated translation is currently sophisticated enough to fulfil a plethora of business needs at a reasonable cost.

Will it work for everyone, though?

To answer this question, we will try to dispel the 3 automatic translation related myths. Read on … this article can make you perceive machine translation in a completely different light.

Myth 1. Google Translator and similar tools produce nothing but hilarious results

What is the general perception of computer assisted translation?

We all know Google Translator. Some of us like to mock this tool, quoting telling examples of what went wrong. For instance, in the process of automatic translation the country of Turkey can become a bird and turkey the bird can suddenly be a member of the Muslim faith.

Hilarious translations are not uncommon and some of them make us raise our eyebrows. For instance, how come Google translated Lady Gaga into Britney Spears from Malay into French? Or how come Android 2.3.4 became iPhone 2.3.4 when translated from German to English?

google-translate-funny-examples

Source: Odee

All the same, humorous examples of translations generated by computers should not be the takeaway here. Even though such translation crimes are factual, the hype they attract distort the bigger picture.

First, tools such as Google Translator are utilised on a daily basis and not … as a joke. Second, in the majority of cases they help to achieve one’s goal: the basic understanding of what the text is about. Third, they’re downright convenient… unless you miss the times when browsing heavy dictionaries was the norm.

Just imagine that you need a translation of a few sentences in a foreign language you don’t understand. You need it immediately. What if there was no Google Translator? Would you translate it manually, word-by-word, using a dictionary? Would you start calling your friends to ask if they know anyone who can speak the language in question?

The fact is that we may take Google Translator for granted. And machine translation for that matter. But think about the era before & after automated translation was easily available: times were much harder before.

Interested in checking YOUR international sales potential on top marketplaces? Book a FREE consultation with our eCommerce expert!

FREE demo

Myth 2. Human translation is always a more accurate option than automated translation

Do you believe that a translation job done by humans is always more accurate compared to the same job carried out by machines? Do you expect human translation to be error-free? It can be in many cases, but think about this ….

… to err is human.

To dispel Myth 2, let us give you a few real-life examples we know from the eCommerce industry.

In our business environment we have been affected by critical errors made by … people. For example, in the past we would use humans to convert sizes from one country to another. We had a lot of critical errors due to human errors, despite the fact we hire professional translators that are well trained.

So what’s so tricky about size conversion? The same dress may be size 8 in the UK and Australia, 4 in the USA and 36 in France. As a rule, translators don’t have to be experts in international size conversion charts.

international-ecommerce-size-conversionConverting sizes can be difficult and confusing for human translators. It’s not a question of maths, but has more to do with the fact that various brands can have various conversion tables. Sizes can also differ between sellers and models.

As a result of this complexity, we needed to come up with an efficient and error-free solution. Eventually we had to automate the process and teach the machine to convert sizes based on brands. This turned out to be a fantastic move: the number of errors decreased significantly.

The solution above illustrates that humans are not always better than machines. In fact, believing that humans can do a better job than machines means rejecting progress. Consider complicated calculations carried out by computers in milliseconds. Or competitions, such as chess games, when computers win against humans.

Machines are there to make your life easier and more efficient. Accuracy is their unique selling point.

Why? Machines can produce more repetitive and consistent content. For instance, if machines are properly programmed with specialised vocabulary in a given product category, they will always work in an accurate and consistent way.

When it comes to human translation, though, the same phrases are likely to be translated in various ways by different human translators. They can also be inaccurately reproduced. In general, in the case of human translation, there is a big scope for inconsistency.

To illustrate the problem think about this … A human translator can specialise in car parts, but may not be familiar with proper fashion terms. This can result in wrong, confusing or inaccurate translations in the fashion category.

The translator’s profile and the specialisation issue is eliminated in the case of machines. The reason being …

… a good automated system will be programmed with what a human being can be potentially confused about. This programming will always result in correct and consistent content.

Myth 3. Humans are not needed in the case of machine translation

So we’ve highlighted that computers can actually be more consistent and accurate than human translators. However, it doesn’t mean that machines should be left to their own devices at all times.

Human expertise is needed to create, improve and control the automatic process.

In eCommerce, we’ve observed that combining human expertise with continuous teaching of the machine to accurately localise product listings is the key to delivering consistent quality.

Hence, our Quality Assurance Team is responsible for regular quality checks and constant improvement of our machines and translation resources. Our specialists validate dictionaries and specialist translators confirm domain specific vocabulary.

Additionally, we have opened our R&D Department in Barcelona. This is where world-class linguists and machine translation experts ensure the best quality of translated items and create the bestecommerce-technology-human-touch translation mechanisms.

The role of our developers cannot be underestimated either. Webinterpret translates from 60,000 to 1,000,000 titles per week. These are big numbers and it may be a challenge to keep up. After all, translations must be ready on time, regardless of the amount of work. Luckily, thanks to our developers, translation assignments and deadlines are kept an eye on and distributed in the most efficient way.

According to Joanna Chodzynska, Product Owner at Webinterpret responsible for the improvement of translation mechanisms…

Every great technology is created by people who are behind it. Thus, what we offer at Webinterpret is not only technology, but people who are passionate about creating a state-of-the-art solution crossing the borders in the machine translation area.

Even the best technologies will not prove their worth without the human touch. At Webinterpret we offer automation, but we recognise the importance of human input, even in the case of computer assisted translation.

So all myths considered …

Is machine translation always the best solution?

The answer may not surprise you…

It all depends on individual needs, the context and the expectations of the target audience.

In some contexts, the mere understandability of the text will do the job. In other cases, the reader will expect a literary text whose primary function is aesthetic. When it comes to the latter, as of today … humans are irreplaceable.

If …

  • the fluency and the error-free quality of the text is important to the recipient
  • the superior and exact translation quality, e.g. in the case of official documents, is needed
  • the reader is demanding and relies on word-for-word accuracy
  • the reader wants to appreciate the literary value of the text

… automation may not be the best choice.

Moreover, machine translation can work better for some language pairs and subject matters. Thus, the level of quality and accuracy can vary. The quality of the source text can play a big role here as well. To minimise the risk of translation issues, the source text should be clear and error-free.

Finally, accuracy is the strong point of a machine, but accuracy can also be .. its weak point. Machines won’t pick up on translation subtleties and express the meaning as a human would. For instance, texts may contain ambiguous words and sentences where the context is crucial. There are simply factors that machines won’t be able to determine.

But here’s the big BUT…

… if we take eCommerce and product listings … automated translation may be the best solution in most cases.

Product listings are not poetry. They’re not official documents. So the tone may be less important. Nevertheless information must be correct. As long as listing information is factual and clear, the automatically translated listing does its job.

What does it mean in business terms? Efficiency, acceptable quality and a lower cost.

eCommerce-automated-translation

In a nutshell

To summarise, machine translation shouldn’t be looked down on as it has numerous advantages from a business perspective:

  • It’s cost-effective: more efficient, quicker, cheaper. Generally, more is translated for less.
  • It’s quick: machines translate content instantaneously. It takes time for humans to translate the same amount of words.
  • A single machine can be programmed to use multiple languages whereas it is rarely the case that one person speaks, let’s say, 10 languages.
  • Machine translation can be programmed to use key terms used in a specific industry. With multiple translators, this can become a pretty … messy process.

The general perception of automatic translation does a big disservice to it, overshadowing the huge advantages of machines.

Hence, we’d like to emphasise the benefits behind automated translation. In the case of international eCommerce, it has proven to be the best solution for the majority of our customers.

Of course, we’re not suggesting that machine translation should replace human translators altogether … who would like to read a novel translated by Google after all! However, when the price, speed and consistency of translation is at play, machine translation is likely to be an optimal solution.

Further reading & other sources

7 must-know facts about the automatic translation of your listings
International sales: the language barrier finally overcome
[INFOGRAPHIC] Overcome the language barrier in international eCommerce
How to prepare your online business for global sales

Related Posts: