So, if your boss were to come to you and say, “You need to do all of that, but it needs to happen fast" then it probably wouldn’t be a good day…
The solution? Enabling your company to automate as many content and translation processes as possible. By understanding the tools that are available in both content and translation management, and how they can work together, then ‘going global’ can happen as fast as the click of a couple of buttons.
All sounds great – but how do we get there? In this post, I’ll walk you through the considerations and options available to you in making ‘going global’ a reality. But first, things first – removing the humans.
Without doubt, going global slows down when people are involved. It simply takes time for humans to do things with information – create it, translate it, publish it and market it. The place to start in speeding your time-to-market is to look at every part of the information chain in your company to see where a computer can do a faster job.
If that sounds like a daunting task, then let me break it down for you. You’ll soon see that there’s an opportunity to automate a process somewhere at each step of the way. Even if you don’t manage to implement all of these technologies straight away, at least you’ll be able to show your boss a clear plan of how to get moving!
You may think that creating content is a human-only endeavour. After all, computers simply aren’t able to write something descriptive, are they? Actually, they are. If you take a look at an entry for a small town anywhere in the world on Wikipedia, you’ll see something like this:
“Kwethluk (Kuiggluk in Central Alaskan Yup’ik) is a city in Bethel Census Area in the U.S. state of Alaska. At the 2010 census the population was 721."
That introduction, along with much of the rest of the description of Kwethluk, was created by a computer. More specifically, using Natural Language Generation (NLG) technology – taking a set of facts and creating a human language description from them.
Smart companies that deal with a lot of factual material are either already using NLG, or looking closely at it – it’s not restricted to Wikipedia and in fact, there are a number of commercial offerings available today.
So, in the search for speed, take a look at the types of content your company is creating. There is usually a type of content that follows the pattern of “facts become words". Indeed, for many companies these factual descriptions actually form the majority of the content they create.
In the travel industry, for example, it is inventory descriptions – hotel descriptions, room descriptions, cruise cabin descriptions, rental car information, and so on. The majority, if not the entire, description will simply be a synthesis of facts in written language form:
The 75 air-conditioned guestrooms offer flat-screen TVs, complimentary Wi-Fi, and coffee/tea makers. En-suite bathrooms include hair dryers and power showers. The property is completely nonsmoking.
These days, you really don’t require a human to create this type of copy.
Translation without the lunch break
If, again, you think the task of translation is a human-only endeavor, then you’re never going to improve your time-to-market. With the right bit of kit, you can automate both translation process decisions as well as the translation of some content itself. The key technologies here are a translation management system and machine translation, respectively.
Translation management systems (TMS) such as our own SDL WorldServer and SDL Language Cloud can take over much of the decision making about which documents need to be translated, into which languages, at which particular level of quality. Again, it starts with identifying the type of content it is; descriptive content, marketing content, legal content, etc. Often different content types reside in different content management systems, or are created by different teams, so the decision about its ‘type’ has already been made!
A TMS can be customized with a number of “workflows" that mimic human-type decisions. The difference is, of course, that such decisions can be made in milliseconds and the computer doesn’t take a lunch break.
Examples of some common translation decisions are:
• All legal content must be sent to a legally-qualified human translator in that country
• This hotel description needs to be translated into Spanish, so it should be translated using a machine first, then quality checked by a human translator
• This PowerPoint slide deck for marketing should be translated into Italian, Japanese and Korean only if the total cost is below USD$80
Decision points can be built into TMS workflows to make such decisions automatically. Indeed, by tapping into other data systems within the company – sales data, customer profiles, and so on – these decisions can become very sophisticated. Here’s an example:
• Translate the description for this single hotel into Spanish using only machine translation if we have sold less than 10 room-nights over the past 6 months, and it is going to be displayed within our late-deals mobile app for the next 24 hours
It’s unlikely this type of decision making would even be possible without teams of humans pouring over spreadsheet reports and coordinating effort across multiple silos (Sales, Content, IT) within the company to deliver a cost-effective, rapid translation. In such cases, not only are you improving time-to-market by automating the process, but you’re creating an entirely new line of product possibilities from your existing inventory.
As for machine translation, I’ve already indicated above that it can be part of an automated workflow. What you may not appreciate fully, is how it can be applied for reducing costs and moving things along that much faster.
Machine translation is actually very good these days. SDL trains its own BeGlobal machines on industry vertical content. This means you have access to machines that have been tuned to create translations specific to, say, travel, the automotive industry, life sciences and consumer electronics. Indeed, some content types can be translated almost exclusively by machines – like the User Generated Content at TripAdvisor.
Machine translation can also be used as a start point in a type of human-led translation process we at SDL call “iMT". Here, the machine is given a chance to translate a piece of content first. A human follows next, either quickly approving the translation if it’s good or editing it if the translation needs a bit of extra work. For many languages, this is a faster process than full human translation alone.
If you want to get even faster, then SDL TrustScore will enable you to automate the decision about how good (or otherwise) the machine translation actually is. You can then automatically have that translation proceed directly to publication, or go and get patched up by a human translator. Because machine translation is often cheaper – and WAY faster – than human translation you should be trying to remove the human from the translation step, wherever possible.
Websites at the speed of light
Having all of this computer-generated and computer-translated content available is lovely, but now you need to get it out to the market as fast as possible. It’s no good getting these steps to happen in a few minutes if it’s going to take hours or days to get it out to your global websites!
The approach that SDL Tridion Sites takes in solving this issue is very clever. It uses a concept called “blueprinting", which is a method of templating the content and layout of your webpages specific to your various global audiences. It’s extremely intelligent in the way that it does this and, again, can make a number of decisions about what to show, when and to whom – a completely automated process. Humans need not apply.
Once these blueprints have been set up, additions and updates can be made across the internet, literally at the speed of light. Moreover, these changes can be made according to the local requirements of each community accessing the sites, since these decision points can be programmed in.
To fully appreciate the power of this approach, you need to contrast it with what happens today in many companies with a global web presence. Each website is fundamentally a separate internet property; each must be updated individually, and the content it receives from HQ must be manually assessed as appropriate for its audience. If you wanted to add a huge overhead in your time-to-market strategy, this would be a great way of achieving it.
The fact that SDL Tridion Sites is integrated with SDL’s translation technologies means that the entire content creation-translation-publication chain can be fully automated. Can you imagine a hotel description being generated in a few milliseconds from a database of facts using NLG, translated into 15 languages in a couple of minutes using machine translation, its quality assessed and published in a fully localised context automatically?
That’s what I meant at the beginning of this post by “clicking a few buttons".
Tomorrow is here today
Time-to-market matters for every company. The threat from global competition, the need for market growth and the expectation of instant gratification by ‘mobile-optimised humans’ means that companies need to continually innovate to deliver their products and services.
The only way to achieve speed with the content describing and marketing your products and services is to automate. Machine-based content creation, translation management and global, location-sensitive publication has been evolving for decades – and is now well tested within enterprises. Even SME’s – and certainly start-ups – should be considering some or all of the approaches I’ve outlined here.
If you want to be in the game, get in touch.