Machine translation puts rocket boosters behind any e-commerce operation. Is your site up to speed?
E-commerce is now the dominant strategy for any business wishing to build an international customer base.
Businesses hoping to take the e-commerce fast track to the global market, however, soon discover there is more to targeting overseas markets than simply providing online access to their goods.
For an e-commerce site to achieve anything beyond superficial market penetration in foreign countries, it’s essential to overcome the obstacle of translation.
The Challenge of E-commerce Translation
Preparing an e-commerce website for an international audience goes far beyond translating the individual product descriptions.
Technical specifications, promotional messages, and FAQs all need to be localized for the target market. The surrounding web content must also follow suit: the home page, terms and conditions, the menus, and every stage of the payment processes—from basket listing to receipt for payment.
There are also the meta tags to consider—the hidden text that usually stays behind the scenes to help the website be found by search engines: the title, description, and keywords.
The entire customer journey must be recreated as if it was originally designed with your target consumer in mind.
If this makes you feel that e-commerce may not be the simple path to world trade domination you thought it was, consider the reward that could be yours once your e-commerce site has been localized for each market: a huge and lasting boost to your sales.
Localizing your e-commerce website is, therefore, a good thing to do—but there’s work to be done if you want to maximize the returns on your investment.
And if this new aspect to the workflow slows down your products’ time to market, the question must be asked: is it worth it?
The Importance of Winning the Time-to-Market Race
While quality is vitally important for ensuring lasting success, there’s something even more important in the competition for market share: speed. The best product does not always win. It is the fastest to market product that wins the race for customers.
Minimizing time-to-market cycles for translation ensures products can be distributed in many languages more quickly and at larger volumes. In contrast, slower time-to-market cycles equate to a later launch date, which results in delayed and reduced sales, as your business loses out to competitors who have acted more swiftly.
Traditional human methods of translation are struggling to keep up with increasing pressure to deliver at ever-faster rates.
So, how can a fast-moving e-commerce business satisfy its customers’ demand in their native language? The answer lies in automated translation, also known as machine translation.
Using Automated Translation to Speed Up Time to Market
In recent years, software solutions have been developed that translate text into another language using advanced learning capabilities that ensure a more sophisticated ‘human’ fluency.
Modern automated translation systems can convert text while recognizing and maintaining a company’s unique style and tone of voice. While humans still play a part in the process, automated translation can significantly reduce the time taken.
Over the years, three dominant methods of automated translation have become available: Rule-Based, Statistical, and Neural.
Rule-based machine translation (RBMT), which emerged in the 1970s, is designed around an extensive set of manually-created vocabulary translations and grammar rules. Although effective, the results of this method tend to be somewhat mechanical in style.
Statistical machine translation (SMT) uses a more sophisticated approach based on existing translated phrases gathered from years of often human-translated texts from large global organizations, such as the UN, WHO, EU, and the World Bank. When presented with a new text to translate, this comprehensive database of accepted previous translations is evaluated using statistical probability models to derive the most logical interpretation.
Neural machine translation (NMT), developed in the mid-2010s, accumulates knowledge by seeking out patterns in source material—much like a child’s brain would—piecing the world together one experience at a time. When used for translation, a neural network uses its acquired knowledge of previously encountered words, phrases, and sentences to determine the most likely interpretation, one word, then one sequence of words, at a time.
Neural networks are designed with vital operational features located at the network’s end points, rather than in central hubs. This makes delivery rates much faster and memory usage more efficient than when using SMT. Sentences are mostly translated fluently, with an understanding of informal and formal language, and the need for post-translation editing shown to be 25% lower than with SMT.
As this system relies on learning by experience, the accuracy of the translation is dependent on the quantity and quality of the data previously incorporated.
It’s Time for Retailers to Adapt
For an e-commerce operation to thrive, retailers must translate and adapt their online content to provide a truly native shopping experience and keep pace with constantly changing customer demand.
Summa Linguæ Technologies is taking this a step further with a system that bypasses the need for translation by starting the process at the content creation stage.
Nordics Managing Director Lea Backhurst explains,
“We take a set of key product features—such as color, size and materials—and automatically generate an original product description in the requested languages, so there is no initial text to translate, all language versions are created in parallel.”
Increasing speed and quality while minimizing cost is raising the potential return on investment and motivating e-commerce retailers to seek ever-smarter solutions.
Humans will always have a role, but there is no denying that, in the battle of man versus machine, it is automated systems that are helping e-commerce retailers win the race to market.