August 2012
By Uwe Muegge

Uwe Muegge has more than 15 years of experience in the translation and localization industry, having worked in leadership functions on both the vendor and buyer side. He has been with CSOFT International, a provider of language services based in Beijing, since 2008, and he currently serves as Senior Translation Tools Strategist for North America.



The silent revolution: Cloud-based translation memory systems

If you are a translator or someone involved in translation, have you ever day-dreamed about a translation memory system that provides all its features without the headaches: a TM system that doesn’t conflict with other applications; one that runs smoothly on any system including a Mac; a translation memory system that you never have to update; and last but not least, one that doesn’t cost you hundreds of dollars just to get started. If this vaguely describes your vision, I’ve got news for you: This type of translation memory system is a reality and has been for a number of years already.

A brief history of cloud-based translation memory technology

In a traditional computing environment, all processing power and data are located on an autonomous local device, typically a desktop or laptop computer. In a cloud-based environment, on the other hand, the local computer serves primarily as an input/output device that communicates with a remote server, and it is within the remote server that most of the processing power and data reside.

Consequently, the term ‘cloud-based translation memory system’ refers to a TM system, where the translation memory software and linguistic assets (i.e. the translation memory database, glossaries, etc.) are hosted on remote, web-enabled servers that linguists access using either a thin client or just a standard web browser. This type of translation tool made its debut about ten years ago, when large translation service providers and large buyers of translation services started deploying web-enabled translation management systems, such as Lionbridge’s Freeway and Idiom’s Worldserver. At that time, only those translators working for one of the few owners of these systems, had access to them.

In the second half of the 2000s, things changed dramatically. In 2007, Lingotek, a newly founded translation technology company, made its cloud-based translation memory system available to any translator for free. But it was in 2009 that cloud-based translation became known to a wider audience for the first time, when Google launched the Translator Toolkit: a free, full-featured cloud-based translation memory system that was primarily designed to improve the translation quality of Google Translate, Google’s proprietary machine translation system.


Figure 1: Google Translator Toolkit was one of the first cloud-based systems available to freelance translators and is one of, if not THE most popular cloud-based translation memory systems.

Today, translators can choose from a wide variety of professional cloud-based translation memory products. The Translator Toolkit probably has the largest user base today and is still free, as is Wordfast Anywhere, the third and newest member of the Wordfast family of translation memory systems. Lionbridge Translation Workspace, Memsource Cloud, Wordbee and XTM Cloud, to name just a few, are examples of fee-based services.

What’s so great about cloud-based translation memory systems?

No application to install

While some cloud-based translation memory systems require users to install a thin client, such as a plug-in for Microsoft Word, many rely entirely on the functionality of a standard web browser to connect to the remote web server. Since in all cloud-based TM systems the “heavy lifting” (e.g. segmentation, TM lookup, and glossary lookup) is done on the server side, users of cloud-based systems don’t have to worry about the involved, multi-step installation procedures characteristic of conventional desktop translation memory products. Typically, all it takes to get started with a cloud-based system is an Internet connection signing-up for the service.

Up-to-date translation software, every time!

Among the most inefficient, and truthfully quite maddening, aspects of working with traditional translation memory products are the required installations, bug fixes, service packs and updates – not to mention upgrades – to keep the system running and in sync with clients and colleagues. As a busy language professional, the last thing you want to do is spend (non-billable!) time on software maintenance.

With cloud-based systems, users don’t have to worry about updating their software as maintenance is performed on the server-side. That is to say, the software vendor takes care of all updates, so users can rest assured that they always use the latest version of their software. A nice feature of cloud-based systems is the fact that updates are not only deployed automatically without user intervention, update cycles for cloud-based systems are also typically much shorter than those for traditional software products, with some cloud-based TM system vendors releasing new versions on a monthly basis.

Translate on a Mac or mobile device? No problem!

In the past, cross-platform support has been a major issue with commercial translation memory products. While it is certainly true that translation memory systems for non-PC operating systems have been available for more than ten years (with Wordfast as a pioneer in this area), the market-leading translation memory systems run only on Windows, even today.

Since cloud-based translation memory systems typically require very little processing power and memory on the user side, these systems support not only computers that run traditional operating systems (e.g. Windows, Mac OS, Linux), but also iOS- and Android-based mobile devices. In the age of cloud-based translation memory systems, translators no longer need to have expensive computers with fast processors and lots of disk space in order to take advantage of the latest translation technology. Any Internet-ready device, including tablets and even smart phones, can now do the job!

Easy collaboration

The benefits mentioned so far indicate that cloud-based translation memory systems are more convenient and easier to use than traditional desktop systems. That alone, in my opinion, would qualify this technology as an evolution of translation technology, but not a revolutionarily new development. However, the following characteristics of cloud-based TM systems are game-changers for freelance translators, as well as for small and medium-sized language service provider businesses.

In a traditional translation environment, collaboration is possible but difficult, because sharing is a process separate from translation. If more than one translator per language wants, or needs, to participate in a project, the document to be translated has to be divided among translators, and the translators themselves need to export, e-mail and import translation memories and glossaries on a daily basis to leverage TM matches and ensure consistency. And even then, there is always a gap between TM updates, during which translators a) cannot benefit from the translations and potential TM matches their colleagues create, and b) create potential inconsistencies or conflicts with their colleagues’ work. In other words, using desktop translation software to share TMs and glossaries among translators who work on the same document is a complex and highly inefficient strategy.


Figure 2: Tà, CSOFT’s free cloud-based translation eco system, which is currently in limited release, features integrated machine translation, completely user-customizable workflow management, and advanced automatic quality assurance functions.


With a cloud-based TM system, where all linguistic assets are stored on a single centralized server, sharing translation memories among multiple linguists is effortless: By granting other users access to a project, either through a simple system setup or via e-mail invitation, teams of almost any size can instantly collaborate. Translators can see and leverage the translations their colleagues create the moment a sentence is entered in the translation memory database. This means maximum productivity, maximum consistency, and no time wasted sending files and performing non-value-added management tasks.

Better yet, for urgent projects, cloud-based systems enable editors and reviewers to work on a document that is still in translation. As all data are stored in a single repository, using a cloud-based translation memory system can dramatically improve translation turnaround time compared to traditional desktop and closed server-based systems. Unlike previous computing environments that force users to structure projects sequentially and manually perform handoffs whenever they want to share information, cloud-based systems give multiple users in multiple roles simultaneous access automatically!

Workflow, project management, and portal functions

Another nice feature of many cloud-based translation memory systems is that they offer more features than most traditional desktop translation memory products. For instance, many cloud-based TM systems support workflow functions that will automatically notify a designated person once a specific phase of a translation project has been completed (e.g. notifying an editor that the translation phase has been completed, or a reviewer on the client side that the editing phase has been completed).

Speaking of clients, some cloud-based translation memory systems make complete translation portals available to their users. In this type of translation environment, clients of freelance translators or small translation agencies can log on to the system and instantly have access to automated quotes for new projects or status information for existing ones, all at the push of a few buttons. These portals can give clients of even the smallest translation business an unprecedented level of customer service without causing the slightest distraction to linguists.

But it doesn’t stop there! Many cloud-based translation memory systems offer additional features in one or more of the following areas:

  • machine translation (for post-editing, performance optimization)
  • instant messaging (to resolve translation issues with a colleague in real-time)
  • quality assurance (through automatic terminology and TM consistency checks and by the integrity of tagging, formatting and numbers)
  • customer management and reporting (e.g. client rankings by revenues)
  • billing and online payment

Low cost solution

Most of the services mentioned in the sections above have been available in web-based translation systems for a number of years. What’s new is the fact that users don’t have to spend six figures on software that demands a high-end IT infrastructure, which requires specially trained human resources to operate. Cloud-based translation memory systems are typically available on a subscription basis, and users only pay for what they use, month by month. With cloud-based TM systems, there are no huge up-front costs and no long-term commitments. And with plans starting at less than $50 per month, even translators just starting out in the translation business can begin using this incredibly powerful technology today.

What’s not to like?

With all of these benefits stacked in its favor, why isn’t cloud-based translation memory technology used by the majority of translation service providers today? One big reason might simply be that most language professionals haven’t heard opinion leaders talk about either cloud technology or the compelling features of cloud-based translation. In addition to not being on the radar screen of many, cloud-based translation memory products also include several features and characteristics that are controversial to some people, causing them to shy away from using such products.

Requires constant Internet connection

While some cloud-based TM systems include an offline translation tool, many do not, which is a concern for some translators. For many in this field, the Internet is an invaluable resource, without which work may become inefficient. Also, some of the early web-based translation memory systems had serious issues with response times – the memories of impatient thumb twiddling while linguists waited for TM matches (or even simply for the cursor to move to the next segment) may have been enough to turn them off from trying out a more updated version.

Fortunately, most vendors of cloud-based translation memory services use cloud-based technology themselves. Instead of running web applications on their own servers, they use high-bandwidth, high-availability cloud-based service providers such as Amazon EC2, Google App Engine or Rackspace. As a result, cloud-based translation memory systems today have typical response times of below 10 ms (which is faster than the blink of an eye) and uptimes are in the neighborhood of 99%, which means that this technology is very, very reliable and will almost always be there when linguists need it.

Privacy issues

Yes, it is true that some of the free cloud-based translation memory systems, most notably Google Translator Toolkit, by default make the translations users enter into their translation memory available to all other users of that system. To have some of their most valuable content be openly searchable on a website is a scary thought for many buyers of translation services, which is why some of these buyers now have clauses in their non-disclosure agreements with freelance translators that explicitly prohibit the use of cloud-based translation tools for their translation projects.

However, it is also true that the vast majority of cloud-based translation memory products by default keeps the translations users enter absolutely private and confidential. Even users of Google Translator Toolkit and other online TM systems can turn their open TMs into private ones, invisible to all other users, by simply changing a preference setting.

So to be perfectly clear about the privacy issue: It is absolutely not true that cloud-based translation memory systems by necessity, i.e. because of their design, make the intellectual property stored in their databases openly available on the Web. Much like an online banking system, the vast majority of cloud-based TM tools, including Google Translator Toolkit, use data encryption for the data traffic between the local computer and the web-based translation server. On a personal note, all cloud-based translation memory systems that I am familiar with keep translation memory data safe and secure, either by default or through one or two mouse clicks.

Control over linguistic assets

Finally, and this is a concern that cannot be easily dismissed, some translators are simply not sold on the whole idea of collaborating and sharing linguistic assets, especially with clients.

It is a reality that some linguists are more experienced than others; this competitive advantage is something some senior members of the translation fashion are reluctant to share with their more junior colleagues for fear of eroding or reducing the higher rates associated with having more translation experience.

It is typically the same type of linguist who believes that translation memories and glossaries are the translator’s most valuable assets and are best kept in a safe place and out of the hands of their clients.

Clearly, translators who subscribe to the views expressed in this paragraph will find cloud-based translation memory systems unpalatable – especially if those translators already own a desktop translation memory system that they are happy with.

Benefits outweigh drawbacks

Cloud-based translation memory systems are the type of translation tool many translation professionals have been waiting for. This new breed of tools offers many of the same features as traditional desktop and server products, however, cloud-based services are much easier to launch and maintain. And, more importantly, cloud-based TM systems allow users to collaborate to a degree that's simply unimaginable with the older type of translation tool. Additionally, some cloud-based translation memory systems enable individual freelance translators and smaller translation agencies to provide the same automated services to their clients (e.g. translation quotes and status updates 24/7) that previously only large translation service providers could implement. And best of all, the cost of using a cloud-based TM service is typically much lower than the start-up cost of a traditional TM product.

While some translators and agencies will not readily embrace cloud-based translation technology for reasons of their own, I think it is a safe bet that five years from now, the vast majority of translation providers and buyers will use cloud-based translation systems to conduct business.

As more translators understand what cloud-based translation systems can do for their business, and as more translation buyers start asking their suppliers to provide the level of customer service only a centralized automated system can deliver, the adoption rate of cloud-based TM systems will begin to increase rapidly.

Further reading

Page 1 from 2
12 >>
#4 Lam Lam wrote at Mon, Dec 16 answer

I used cloud-based systems and as a translator my objection is because the clients thought they can pay my income less on the ground that this is post-editing work which turned out to be as or maybe more demanding and time consuming as normal translation without the system. If you really work on it, you know that this system is, in its essence, no more than a kind of Google translation which will churn out messy results. Working on such messy results, we translators have to erase the whole thing and write it back from scratch if we dont want to have wierd sentences. We cannot share linguistic ideas when underpaid and mistreated by such clients. Peiod.

#3 LocalizationShop wrote at Sun, Aug 12 answer

Great article. It nicely captures the essentials of working with cloud based TMS and the fact that there is nothing to worry about. Once people get use to it, it will help increase efficiency, speed, accuracy and reduce down time. Thanks for sharing!

#2 Andrzej Zydroń wrote at Mon, Aug 06 answer homepage

Thank you, Uwe, for such an informative article. As you know I am also very enthusiastic about Cloud-based TMS, which I consider, like yourself, the solution to many of the interoperability problems that dog translation workflows, add costs and drain productivity. An important note is to emphasize the need for standards adoption, such as those embedded in the OASIS OAXAL reference architecture. We at XTM-INTL value your comments and wholeheartedly agree with you.