For large-scale globalization and localization projects, translation memory (TM) is essential. Think of it like a record of every translation you’ve ever done: when new, untranslated strings (translation units) come up in your queue, this archive searches for possible matches from translations previously done. The tool can also find “fuzzy matches,” which returns phrases and sentences that are only partial matches.
The result? Lower costs and improved translation speeds because you’re re-using translations instead of paying for the same phrases and sentences over and over again.
The problem, however, is some vendors want you to pay for the privilege of a translation memory download, and may even claim your translated data is theirs to hold.
Who Owns Translation Memory?
Does possession grant a provider ownership over your translated content? Absolutely not. Just as untranslated content belongs to your brand, even the finished product should be yours to keep. Steer clear of any company that thinks otherwise. Ideally, you should receive a set of translation files at the end of each project containing both source and target language content. These are often supplied in translation memory exchange (TMX) format.
If you use a cloud-based translation management system, you’re able to create a long-term TM solution that allows translators to collaborate on specific projects and enable real-time updating of translation memory. But even in this case, check if you’ll be charged for storing and accessing the TM files in their cloud.
Why Work with Cloud-Based Translation Platforms?
So what does true cloud-based TM look like? First, it should include multiple translation entries for a single source string—one for each translated language. In addition, translation memory tools should automatically scan new content for possible matches and complement human translators‘ work by providing suggestions for the matches. This allows you to skip translating the boilerplate text of your press release or brand slogan over and over.
But you should also be free to do more: for example, importing your own translation files such as those from iOS, Android, gettext, yaml, java properties, XML, and JSON. In other words, if you choose to go purely in-house for a given project or incorporate older translated content you have lying around into a TM solution, it should be no problem. By the same token, downloading translated content should be simple, straightforward, and free.
This means no complicated access requests. The ideal translation software makes it easy to export some or all of your TMX files to the desired location. This also makes it easy to pack up your translations and take them to a new provider or move them back in-house. Lastly, cloud-based solutions should make it easy to share content among stakeholders and human translators to ensure projects are on time and on budget.
Translation memory is a crystal-clear record of every localized piece of content you’ve created, allowing you to save money and time through complete and fuzzy matches. It also contributes to quality hugely, by ensuring previously approved translations are used and new errors prevented.
Words are power; your words—and their impact—belong to you in any language.
Image source: BigStock