Translation Memory Software 101
With Transifex’s Translation Memory software, your translators and the whole localization team can skip over repeatedly translating the same phrases. This effectively saves you both time and money while making localization easier.
Wanna learn more about Translation Memory (TM) and how you can use it? Then feel free to keep on reading!
Table of Contents
Translation Memory Software & How it Works
TM is one of our many tools that you can use to make localization easier.
Simply put, TM stores your translations, be it for words or phrases, and when it comes across another phrase that is 100% identical, it’s going to automatically fill up that phrase for you.
For example, if you have three identical greeting messages on three different pages on your website, and you translate one of them, TM is going to automatically fill up the other two. Assuming that you have TM Fill-Ups enabled, of course.
But it’s not just about accuracy. Context and human interaction can still play an important role, if you want them to. More about that later.
Keep in mind that you don’t have to set up anything. TM works in the background, and whether you are going to enable some of its additional features or not is completely up to you!
Here is everything that you can look forward to!
Translation Memory Killer Features
This is the core of Translation Memory. By enabling TM Fill-ups, you allow the TM database to automatically “fill up” any translations that have already been completed before.
You can find that option under:
- “Project of your choice”
And make sure that the “Fill-up” option is ticked. Disabling TM Fill-up, on the other hand, means that you’ll be preventing the Translation Memory software from automatically filling up translations that have already been done.
One common reason for disabling TM Fill-Up is because you may want to review all of the available suggestions before adding the final translation. After all, TM suggestions are always going to stick around regardless of whether you have TM Fill-Up enabled or not; more about that later.
This is a tool that’s going to be keeping track of translations and whether there are any matches of 60% or higher.
TM Groups allow you to use the same TM database for multiple projects.
See, by default, Translation Memory stores data per project. So, you are going to have different TM data for different projects – which is how it should be.
However, there are some cases where two projects are identical, but for a different application. For example, maybe you are translating the same app for both Android and iOS, but you have different projects for each.
In this case, since we are talking about more or less the same localization project, it makes sense to use the same Translation Memory database for both projects.
You can find TM groups under the organization settings in “Translation Memory”. To add a new group:
- While in the organization settings, click on “Translation Memory”
- Add Translation Memory Group
- Insert a group name of your choosing
- Add all the projects that should be sharing the same Translation Memory database
- And click on “Apply” when you are done
Keep in mind that each project can only have one TM database, though.
Leverage and Suggestions
As mentioned above, TM automatically fills up identical source strings that have already been translated. However, that only applies to 100% matches. If a phrase is even 1% inaccurate, then that’s where TM suggestions come in.
For example, consider that you are dealing with three similar, but not exact phrases:
- “Welcome to our website”
- “Welcome to our blog”
- And “Welcome to our webinar”
After you translate the first phrase, our Translation Memory software should recognize that the other two sentences are similar, but not 100% identical. Thus, instead of an automatic fill-up, you should get a suggestion in its place.
Even partial matching can greatly reduce how much time and money you spend on a project. But the exact percentage of that can greatly vary depending on the language and even the translators – which is where TM Leverage comes in!
To check your TM Leverage, go to:
- “Project of your choice”
- Resource of your choice
- Language of your choice
Furthermore, you can also check your TM Leverage under the “Reports” section and the localization activity overview. This is a handy tool that you can use to get a better understanding of just how much you can save by relying on TM at the current stage of your project.
Translation can be a tricky thing; localization even more so. A single word can have multiple meanings depending on the context.
For example, “Γεια” in Greek can be used as a greeting, farewell, salutation, or you can even substitute it with “Cheers” when in a feast. One word, four different meanings.
In this case, even a 100% TM match is not going to be correctly translated. You can solve this by making Translation Memory take context into account.
This option is under the:
- “Project of your choice”
- Translation Memory Fill-up (Needs to be enabled)
- Translation Memory Context Matching
You can very easily add context to phrases by using the editor while translating – which is optional, but, in such cases, context can make quite a difference.
When localizing, you usually have both a team of translators and someone who reviews the translations. This workflow is integrated into our platform and you can use it to ensure that the translated content is up to your expectations.
If you wish to, you can integrate that part of the workflow into the Translation Memory, meaning that only reviewed translations will be allowed to enter the TM. To toggle that option, go to:
- “Project of your choice”
- Quality Assurance
- “Allow only reviewed translations into the Translation memory”
That’s a great way to make sure that your TM database only contains clean, high-quality translations.
How Translation Memory Can Help you Localize More Efficiently
Translation Memory Software is great for eliminating the tedious task of having to translate the exact same sentence over and over again.
To be more specific, here is a summary of how you can use each feature for your benefit:
- Fill-up: Have our software automatically complete identical phrases that you’ve already translated so that your translators won’t have to waste extra time and money in order to do it
- Groups: Share a single TM memory with two or more projects that are similar to each other, such as a single app for two different operating systems
- Leverage: Get an overview of just how much TM is offering at the current stage of your project. Increasing your TM leverage means reducing the time and effort that translators put into their work and thus, reducing your overall expenses, as well
- Suggestions: You often get phrases that are similar but not 100% identical. In which case, you can still get partial translations that help with consistency and context while reducing the overall time and cost of translations
- Context matching: When setting up a project, you can add context to words and phrases so that the translators can get a better understanding of what they are translating. TM can also benefit from that to further increase translation accuracy
- Quality assurance: After completing a translation, someone also needs to review it to ensure that it meets certain quality standards. By “forcing” TM to only accept reviewed translations, you can further increase the overall quality of “automated” translations
To get the best results, make sure that you are maintaining a delicate balance with how you use our tools. And keep in mind that the ideal balance can change depending on your translators, language, and industry background, as well.
How to Properly Use our Translation Memory Software
Using TM tools is all about maintaining a delicate balance. If you disable everything regarding quality control, you’ll get tons of automated translations, saving you tons of spendings and time in the short term. However, that may negatively affect quality in the long term.
If you enable every quality control tool, on the other hand, you’ll be no doubt increasing the overall quality of your translations, but you’ll also end up getting a lot fewer fill-ups. Plus, the localization process is going to be significantly longer.
So, it’s really all about experimenting a little bit with the settings and seeing what yields the best results.
You can go ahead and disable or enable every option, but do keep in mind that this won’t be offering optimal results.
Last, but not least, don’t forget that you have the option of deleting TM suggestions. And you can do that either directly through the suggestions or through the project settings.
You can use this option to get rid of repeated suggestions that you don’t find useful. In extreme cases, you can also wipe off the entire TM by downloading it, editing it, and then overwriting your old TM with the edited one. If you are interested in that, we have a step-by-step guide on this page.
Translation Memory VS Machine Translation
A somewhat common misconception is that people often confuse Translation Memory with Machine Translation tools, such as Google Translate. But the two of them have little next to nothing to do with each other.
Translation Memory Software relies on the translator’s work to automatically fill-up or suggest translations. It builds a database from direct human inputs that it can use to accurately predict what else it can automatically translate.
Machine Translation, on the other hand, uses a very complicated and sophisticated AI to automatically translate everything. The problem is that, unlike TM, Machine Translation is nowhere near as accurate.
And while we’ve seen huge advancements in it, most people are still not willing to trust AI tools for accurate translations.
In short, Machine Translation handles everything automatically with mediocre accuracy while Translation Memory is accurate, but it can’t do anything on its own.
How Does Translation Memory Help With Cost Savings?
Translation Agencies often offer discounts on a cost-per-word basis. The industry standard is:
- 50% discount for fuzzy matches (75%-99% match)
- 75% for a 100% match
And that’s not all. With a solid translation database and plenty of 100% matches, translators won’t have to work on translating hundreds, or maybe even thousands of sentences and words; words that you won’t have to pay for.
What is a 101% Match?
As mentioned above, TM doesn’t simply translate based on how similar two words or phrases are. It also takes context into account.
If a translation is 100% match while also sharing the same context, then that’s a 101% match – which is hardly ever going to be a wrong translation.
How Much of a Difference Does TM Really Make?
To speak with numbers:
- 52.3% of our customer’s translations are auto-populated by Translation Memory
- 40% of translations have a 74-99% match/leverage
This can effectively reduce a 2-hour job down to 30 minutes while offering at least a 20% discount for each translation project.
But, of course, your mileage may vary depending on various circumstances.
That’s about all you need to know about our Translation Memory Software. If you need any help, feel free to contact us either through the bot or via email for more information.
Alternatively, you can also try Transifex for 15 days completely free of charge; no credit card required!
Do also feel free to watch our Translation Memory Webinar where our Head of Product, Maya Toutountzi, guides you through everything you need to know about Translation Memory.