backstage   All Products  |   Support  |   Search  |   microsoft.com Home  
Microsoft
  Home  |   Events/Training  |   Subscribe  |   About Microsoft  |   US/Worldwide  |   Downloads  |   MSN.com  |


schematic

Microsoft
Backstage


Operating
microsoft.com


Inside Two
Sites


Solutions/
Best Practices


White Papers:
Windows 2000
Windows NT

More Site
Information


Backstage
Archives




Other
Microsoft IT
Information


TechNet

MSDN

Application
Showcase



English is Just Another Language for Microsoft's Web Localization Team

Continued from front page:

To put it all into context, here are some statistics:
  • Last year more than 24 million words were localized on the content Web sites that we manage, including Product Catalog, Solutions, Direct Access, TechNet, and even Backstage.
  • More than 400,000 words were processed just for Backstage, which is translated into 24 languages.
  • In addition, nearly 1.3 million words were translated for e-mail newsletters such as Exploring Windows, Office News Service, the MSDN Flash, and the TechNet Flash.
  • Another 700,000 words were localized for more static projects (aka "one-time" localization jobs). For example, strings for Web services such as Search and Profile Center.
In the localization world, it is common to have English come first, and then all other languages localized based on the English model - a tedious way to translate. As we reported two years ago (see "Taking A Global View: Localizing for the Web"). It was once common for our localization process to start only after design of the Web site for the U.S. market had been completed and shipped. Not only did this ensure large delays between U.S. and international releases, but it also required resources for re-engineering the templates and editing the content to remove any U.S.-specific content. Now sites are designed to be global from the start. Even double-byte languages, such as Chinese and Korean, and bi-directional languages, such as Hebrew, are accommodated as easily as English.

Philosophy
With challenges like these, we need a consistent philosophy to survive - better yet, thrive - in the ever-changing world of localization, something we have been perfecting in the last eighteen months. Simultaneous shipping of content to 47 subsidiary sites (Microsoft's presence in countries other than the U.S.) and the implementation of database-backend Web tools have made localization more efficient; within this model we are able to provide the subsidiaries with instant access to the latest technologies and a content workflow that is much more streamlined.

Our philosophy of localization has not changed. First of all, process is the key to localization success and to a smooth and efficient workflow. Second, we never forget our customers, both internal and external, the subsidiaries and the end users. Third, we work closely with the product planning, program management, development and test teams to make sure that they understand the requirements of worldwide programs and how to design projects with a global perspective. Our team knows and cares about the needs of the international Subsidiaries.

For example, when designing the registration system for microsoft.com, the Localization Program Manager worked closely with the program management and development teams to ensure that address forms on the site were regionalized, international requirements were included in the privacy statement, and that the architecture of the system allowed for different character sets and country-specific customizations. Planning and cross-team cooperation of this nature occurs before localization begins.

And last but not least, we get and keep great localization vendors who localize Web content and applications both quickly and carefully. They are critical to our success.

Process
We have implemented the best tools to create great process. In May of 1999 we launched version 1 of our Product Catalog publishing tool which includes a worldwide localization workflow, and we were pleased with the results. The Unicode support in Microsoft SQL Server has taken the place of more antiquated approaches to localization (see Figure 1).

Figure 1. Unicode strings for different languages displayed from a SQL Server database.Figure 1. Unicode strings for different languages displayed from a SQL Server database. (Click to enlarge)

On top of the database, we constructed a workflow facility that enables subsidiaries and vendors to input and localize content applicable to their markets, complete with a queuing schedule. The process also lets them know when they are next in the queue to input content to publish. We were very excited about this new tool because we no longer had to deal with a manual workflow.

However, there was one key challenge despite the excellent progress we had made. We discovered that it was not good for everyone. Some of our vendors found it difficult to log on to our intranet and - because of slow overseas connections - very expensive. We decided that we needed to change the process in order to accommodate our customers, both to keep them happy and the cost low. So we are creating a method for them to download the files and work on them offline.

Tools
Out of all this effort and learning, of course we are getting smarter about what it takes to build and implement truly worldwide tools. The next version of our worldwide localization workflow tools will have a crystal clear UI within which our subsidiaries worldwide will be able to continue to choose what they want localized and input their own unique content.

For example, we will extend the Product Catalog publishing tool model to other sets of content, offering our subsidiaries a list of content and they select what they want to have translated. They also have the ability to add their own local market content, and then everything that needs to be translated goes to the outsource Localization Vendors located in-market. Our goal is to encapsulate everything into an XML file and put it into a standalone third-party localization tool. (For more information on the evolution of our catalog tool, see "Building a Better Product Catalog: Putting SQL Server 7.0 and XML to Work in a Three-Tier Solution".)

We are now moving toward using a localization tool with a "translation memory" engine that keeps a record of all content that has been previously translated. This enables us to accurately determine how much content is truly new and how much can be recycled from previous projects whenever we start a new round of localization. We estimate that we can reduce the volume of localization work by up to 25 percent in this manner. For example, a string such as "The following is a list of frequently asked questions" can be put into memory and reused thousands of times. It makes no sense to reword it or translate it again. In addition, translators can use the tool to quickly check how a specific word or phrase has been translated in other contexts, which is crucial to ensure consistency across the many properties on which we work.

In fact, we currently have a pilot program for the Search Web Service, one of our smaller projects. In this program, the strings are exported as XML, dropped into the localization tool, translated, and then returned. We want to see how the platform for XML needs to be tweaked so the localization tool can pick it up without any problems. Once this localization-friendly XML format is perfected, we can use it for other projects including our worldwide localization workflow tools.

Relationships
Everyone on the Microsoft Web localization team works to support the subsidiaries and what they need from the site. We serve as a channel between the subsidiaries and the corporate team in Redmond and work with them on projects such as our worldwide registration system, application string localization, requirements for new global releases, and other globalization and localization priorities. Our relationship with the subsidiaries is a two-way street: We respect their local market and industry knowledge and try to learn as much as we can from them. This, in turn, adds to our success.

For example, in May and June, our localization team hosted Platform Summits in Asia and Europe with our subsidiaries. We shared plans, brainstormed on new ideas, and cultivated our "virtual team" approach for two days, resulting in a better partnership to the shared challenges of localization and globalization.

The other part of our job is our relationship with the localization vendors. These outside companies operate from their own offices, most typically "in country" to the Microsoft subsidiary. They not only provide the translation work but also offer a local market expertise taking the work beyond only translation. Our main connection with the vendors is a project manager (PM). For example, the project manager at the Spain vendor, a native speaker, works in Madrid and employs local translators in Spain. This project manager and his team are able to provide us with crucial insight and culturally specific conventions in language and style to make the site truly Spanish for the local Spain market. Because much of our job is a thorough evaluation of the quality of the translations they do mapped to the overall goals of the project and needs of the local market, our relationship with the PM is crucial.

Before we hire the vendors, they must take a standardized test so we can ascertain their level of expertise. Typically our project size is smaller and we have a faster turnaround time required, consequently we need a vendor who works quickly and who can speak to the local customer. This is vital for localizing content.

It is critical to maintain long-term partnerships with vendors. One reason is that the learning curve can be steep, so it is often in our best interest to keep these relationship vital. If the subsidiary is happy with the localization vendor, we want to continue to do more projects together, resulting in higher quality content.

We also serve as an intermediary between the subsidiaries and vendors. If a subsidiary sees the quality of a vendor’s translation as sub-standard, we establish communications with the vendor to determine: Is it the quality of the content, or the code, or could it be something else that is affecting the translation? Sometimes it could be something relatively simple such as unclear project instructions. We work with both the subsidiary and the vendor to develop and maintain this relationship. Maintaining high quality for each project is our job.

The Microsoft Web localization team ensures that worldwide features are built into all Web applications. With the innovation and utilization of Unicode, SQL Server, and XML, all languages become equal and we can work with English, or German, or Chinese, or Hebrew, or Japanese. For us, English truly is just another language.

Microsoft's Localization Workflow for the Web
Here is a typical workflow for our localization process:

1. The Localization Program Manager (LPM) works with the development team to ensure that globalization and localization are "baked in" to the project specifications from the beginning and not as an afterthought. Development codes strings and/or content so that it's easy to extract into a localization tool. The corporate team peruses the file, makes necessary changes, and then hands this off to the Localization Program Manager.

2. The LPM "pre-localizes" the file by performing quality assurance on the strings, ensuring that everything is included and clearly outlined for the localization vendors, and then sends the final file - edited, frozen and ready for localization - to the vendors with instructions and a deadline. The LPM works closely with the vendors to answer any questions about the project and/or files, such as what a particular string means, what the right context is, the right screenshots of pages, and so on. The LPM also makes certain that the vendors finish the project on time.

3. The vendors send back the localized versions of the content to the LPM who does quality assurance, on the now 29 files (instead of just the one English file!).

4. The LPM sends these 29 files to each subsidiary for review, along with similarly "pre-localized" English versions customized for the non-US English sites. The subsidiaries examine many things including functionality, linguistic cohesion, and images, as well as local-market suitability. The LPM makes sure that the subsidiaries are happy with the vendor’s work and will collaborate with the vendor to make any requested changes based on subsidiary feedback.

5. The subsidiaries may also provide the localization team feedback and, if necessary, open "bugs" in a centralized bug database. The LPM works with the development and test teams to ensure that all bugs are resolved.




© 2000 Microsoft Corporation. All rights reserved. Terms of Use.
Last Updated: September 1, 2000