This is a paper presented by Vladimir Zhdanov of Alibaba Inc. The contents of this Paper are presented to create discussion in the global marketing industry on this topic; the contents of this paper are not to be considered an adopted standard of any kind. This does not represent the official position of Brand2Global Conference, The Localization Institute, or the author’s organization.
It’s 9am. Your project manager just sent an emergency banner request for a smartphone promotion. The content is simple – header, sub, discount, date – but she needs 5 languages in 20 minutes. Even if you could produce quality content with such a short timeframe, your translation agency will take 2 days to generate the other languages. And your designer needs time to integrate the multi-language content and have it reviewed. You let the project manager know – she reluctantly delays the project for 3 days.
The following week, a category manager for home appliances needs to launch a landing page for an important new seller. The content is very similar to the smartphone promotion – discount, date – but, because the previous content was not stored, your team creates it again from scratch and has it re-translated by the agency. 36 hours later, your designer finishes the page and sends it to the category manager.
The next day, your project manager asks for 5 banners for 5 different consumer electronics promotions by end-of-day – in all languages.
Clearly, the demand for native content with quick turnaround is a recurring challenge. You consider machine translation (MT) but, while cheap and scalable, it cannot provide the level of quality demanded by your business. Furthermore, most publicly available MT engines are designed for general-purpose translations – they lack contextual awareness for the content’s specific application.
You could blame your business team for a lack of planning or your translation agency for a slow turnaround, but the fact is it’s 2017 – the industry is changing rapidly, and your team needs to be able to provide quality content targeting multiple markets around the world at a moment’s notice. If you want to survive, you need to keep up.
I work at Alibaba Group in the International User Experience & Design (UED) department’s Content Strategy team, primarily focused on our global B2C marketplace, AliExpress.com. As a promotion-oriented platform with buyers in markets as varied as Russia, Western Europe and South America, we are constantly challenged by tight deadlines, rapid changes and the need to provide quality content for numerous markets.
Indeed, our mobile app supports 18 languages, including the “traditional” EFIGS localization set, some Middle Eastern languages like Hebrew and Arabic as well as East Asian languages (Korean and Japanese). Whenever we create a global promotion such as 11.11, therefore, there is tremendous pressure on our workflow and turnaround time. And as part of a Chinese internet company, we must react quickly to changes in terms of both execution and strategy.
I believe that our experience and learning will be useful for any team facing similar challenges.
So what is our solution? Templates.
Our content follows the Pareto Principle – the majority of our results come from a relatively small amount of highly creative, high-value content. The rest of our requests are high-volume, low impact – such ‘daily projects’ may not require extremely creative copywriting, but they do demand quick turnaround. By analyzing and categorizing our business requests, we are able to break them down into context-specific use cases, and build templates which can satisfy the business needs for the majority of requests. We do this by identifying the content elements needed for each case and creating a ‘content library’ of pre-created, native-level variables which can be utilized for each element.
Guidelines are written by humans for humans. Only native speakers of the target language can fully understand and implement them. Having guidelines without language skills (as is the case with our designers, project managers and content strategists for languages other than their native language) does not give you the ability to modify or create content in language(s) that you do not speak fluently.
Templates are guidelines written by humans for machines or machine-like applications. A well-written template does not require native language skills to utilize. Templates should be written in a user-friendly, intuitive design in order to prevent misuse. This is similar to programming computer code, but from a purely linguistic-orientation.
Let’s look at an example. Imagine that you often need to create banners that mention the date of a promotion, like ‘Discount starts on May. 30. Get ready!”
A proper translation into Russian, according to company’s guidelines, would be “Скидки начинаются 30 мая. Готовы?“
But, if you try to use machine translation, it will give you “Скидка начинается 30 мая. Приготовьтесь!”, which not only fails to follow the guidelines, but also translates the word “discount” in singular “скидка“ instead of plural “скидки”.
Now let’s assume you’ve pre-created a template with all the content except the variable, which would be the date (May 30). It would look like this:
Discount starts on %s. Get ready!
Surely machine can handle something as simple as a date, right? Not necessarily. If you are not careful, MT could leave you with this:
In this example, the tiny detail of the capitalization of the first letter leads you to a completely incorrect translation (“может 15” in Russian does not make sense and means literally “can do 15”).
Instead, it would be wiser to prepare a template with pre-translated date variations:
With this textual data prepared in accordance with your guidelines and use cases, you can re-use it any time you want without requiring additional linguistic resources.
Translations for the ‘content library’ or ‘database’ should be done by native speakers who understand your requirements and guidelines. The raw information can be integrated into your CMS tools, or even into an Excel spreadsheet:
There are several steps you should take to fully utilize templates in your workflow.
First of all, analyze what content you already have and what texts you are producing on a daily basis. Start with just your main working language. Possible places to look for such content includes banners, product titles, push messages — any high-volume, recurring text.
Understand which areas are the priorities so you know where to invest your time, money and resources.
Turn sentences or other elements into templates. Determine which are the static and variable parts of the element. Re-write the static part to make it as compatible with a range of variables as possible.
Example: Most of the time it’s easy to get rid of ‘(s):
You have 5 coin(s) → Your coin balance: 10
Since the variable here is a number, the template would look like this:
Your coin balance: %s
When your templates are ready, translate them with whatever option is common in your company — in-house or outsource. Don’t forget to include a few samples from the template with your translation request.
Ask the translators for feedback. It’s possible that the structure of your template is not suitable for some languages and may need to be modified.
The message “Looking for %s”, works fine in English if you replace %s with a product category name, like “Looking for Smartphones?”.
But in Russian, category name capitalization should be modified to fit grammar rules:
Ищете Смартфоны? → Ищете cмартфоны?
That means the variable has to be modified. It could be achieved either through storing a pool of prepared variables or via automated procedures (e.g. changing capitalization).
Create a database for the storage of your templates. This database should be accessible by partner teams. Following simple instructions, they should be able to retrieve and deploy content from this database instantaneously.
If you have back/front-end developer resources, they could help you with scripts or tools to achieve your goals. If not, don’t worry – as you can see from my basic example above, even Microsoft Excel or LibreOffice Calc can do a lot of automation for you.
In Excel, you can create a sophisticated and customizable template that will suit specific needs by using basic functions such as IF and CONCATENATE:
In an age when Google robots are able to write news articles and NLG is on the rise, templates might lack the futuristic sheen of their headline-grabbing alternatives. But in the real world, not every creative content team has an R&D department with dedicated developers and researchers. In fact, in our experience the majority of them have limited access to developer resources.
Templates are something that teams of any scale can deploy quickly and get immediate returns on investment. It’s also important to mention that engines like Articoolo are focused on one language, while you may instead require multiple languages (18, in our case).
We found certain areas to focus on for templatization – banners and push notifications. We’ve used our internal resources and outsource translations to prepare templates in multiple languages for these channels.
We’re looking into the various applications for AI, and of course our ultimate goal is to allow computers to do most of the bulk content work. The current bottleneck we face is our wide range of languages – most of our promotions require at least 5 languages, and sometimes as many as 18. We must strike a delicate balance between localization, scale and efficiency. With the resources at hand, therefore, we have found templates to be an effective tool to help us achieve this balance and meet our business needs.
- Analyze and categorize your content and requests
- Locate opportunities to templatize content (look for high-volume, recurring elements)
- Anticipate possible misuse of your templates and try to design them with such risks in mind
- Don’t worry if you don’t have developers – Excel could be powerful enough
- Think bigger
Vladimir Zhdanov is a Senior Content Strategist in Alibaba Group’s International User Experience & Design department. He introduced the team to modern localization principles, such as computer-aided translation, and is now responsible for tools and workflow in the translation and localization areas of written content.
Vladimir graduated from Tomsk State University (Russia) with a degree in Linguistics and Translation, and has 7 years of experience in the translation and localization IT industry, including localization agencies and game development companies. He is passionate about making the content process ‘smart’ and implementing modern technology in everyday life.
If you are interested in learning more about the Global Digital Marketing and Localization Certification
please click here. The program offers dual credentials, with a Certificate from the University of North Carolina Wilmington and a Certification from The Localization Institute.