The Critical Role of Translation Memory

Introduction

For brand consistency and efficiency, many organizations re-use specific phrases throughout their websites. But translating this content everywhere it appears gets expensive, fast. Translation memory eliminates this problem.

How Translation Memory Management Works

Translation memory is a common capability of most website translation services. It’s a database that contains every phrase that’s been translated for a localized website. Each phrase is translated only once by linguists.

Once stored in translation memory, this content can be re-published hundreds—or thousands—of times at no additional cost. Translation memory can also be leveraged for omnichannel content.

Benefits of Translation Memory

Speed: Get new translations completed and deployed quickly for frequently used content, without duplicating translation work. When leveraged properly, it’s instant, automatic website translation.

Consistency: Ensures translation accuracy throughout a localized website. Technical details, brand voice and cultural nuances are preserved.

Cost-Effectiveness: Translating content once and re-using it over and over—even months or years later—saves on translation costs and eliminates duplicate effort.

Website Change-Detection Crawlers

A web crawler, sometimes called a spider, is a technology that systematically analyzes, identifies and classifies a website’s content. Spiders are often used by leading translation vendors to detect the amount of translatable content within a website.

Once the spider identifies this content, the vendor’s content-parsing technology is used to splice pages of content into easy-to-translate segments. Once parsed, segments are processed through the translation memory database.

This detection technology also operates on an ongoing basis, after a website has been translated and published. The spider regularly crawls the site, looking for new or updated content to translate.

Any new untranslated segments are flagged for translation. Once these new segments are translated, they are published on-site and stored in translation memory for future re-use.

With this technology, you do not need to create workflows to send a vendor your net-new translatable content. Instead, it is automatically and systematically detected by the vendor’s spider and queued for parsing, and translation, if needed.

Look for Technically Mature Parsing Technology

All website translation services use a content-parsing algorithm of some sort, but not all are created equal. Great parsing algorithms generate high-quality segments for translation.

Look for vendors with parsers that can granularly customize the scope of your website translation project. This will save you money.

Customizing your scope ensures that your project:

• Better aligns with your business needs
• Optimizes the level of translation required
• Gets the most value out of a limited translation budget

For instance, great vendors can block certain segments to prevent the translation of website pages or sections that aren’t mission critical. Sections that feature less important content can be flagged for low-cost machine translation, too.

In addition, great parsers can be optimized to increase segment repetition, which further reduces translation costs. Product descriptions that follow a template are a common example. This content often appears as sentences or headlines with predictable structures across multiple pages—such as a product type, the size of a product, or a product’s color.

Website translation services with less technically mature parsers treat every single variation of these phrases as unique segments that require translation. This can send translation costs sky-high.

By contrast, robust parsers can be configured to recognize these patterns, and instruct linguists to proactively translate this template-like structure and all of its contents (such as all the variations of a product’s colors, stored in your product database).

The content-detection technology can then be configured to ignore these “templatized” phrases on an ongoing basis, since they’ve already been translated.

Repetitive structures can also be commonly found in meta-descriptions. In these cases, these structures are word-for-word repetitions, with the only difference being a use of double quotes versus single quotes, and the presence of HTML in the meta-description.

Mature parsing technologies knows this is the exact same content and standardizes the algorithm to optimize pickup of the virtually identical text for translation.

This ensures these segments are never identified for translation again, reducing translation costs. The translated meta-descriptions provide a bonus SEO benefit.

Optimizing the algorithm for dynamic content is another key differentiator for website translation services. Most websites use dynamically loaded customized content, such as welcome messages that greet users by name, or content based on a user’s location. This personalized experience is great for customers but creates a problem for translation vendors.

The translatable strings of this dynamically generated content hide in application code or databases, which flummox most vendors.

To compensate for this lack of expertise, most website translation services charge their customers to translate a dynamically loaded phrase translation every time that segment is generated. This sends costs through the roof.

Reputable website translation vendors charge only once to translate content, no matter how many times it’s used online or off-site. You can continue to deliver personalized online experiences to your multilingual customers without worrying about skyrocketing translation costs.

Read more about website translation in our ultimate guide to website translation.