What Is Translation Memory and How to Build It? (2024 Guide)

Translation memory has revolutionized the translation industry by enabling translators and language service providers (LSPs) to increase the speed, quality, and cost-efficiency of their projects.

Pro tip: Maximize the effectiveness of your localization efforts with Centus’ translation memory. Create, modify, and manage your translation memory entries from a single dashboard. Learn more.

What is a translation memory?

Translation memory (TM) is a database of previously translated sentences, paragraphs, or phrases and their corresponding translations. The TM software suggests previously translated content to the translator, thereby maintaining translation consistency, reducing its costs, and increasing its speed.

Translation memory vs. glossary, termbase, and style guide

Let’s clear up the confusion around these four terms.

• Translation memory is a database of previously-translated material.
• Glossary and term base are just two terms for the searchable database containing terminology, acronyms, or feature and product names.
• Style guide is a set of rules on the text's style, tone, language, formatting, and register.

Translation Memory IS NOT Machine Translation

Translation memory and machine translation are mistakenly used interchangeably because both technologies aid the translation process.

• Translation memory is a database of previously-translated words and phrases.
• Machine translation is a computer assisted-method of translation, like Google Translate.

Source: Google

How Translation Memory Works

Here’s how translation memory works:

#1 Database creation

Translation memory software breaks down source text into translation units called segments, and saves them in a database along with their equivalents in the target language.

#2 Database search

During the translation of a new text, translation memory software automatically searches the database for matching segments.

#3 Suggestion of matches

Translation memory suggests perfect and fuzzy matches to the translator. Perfect matches are identical to the source segments. Fuzzy matches, on the other hand, only partially correspond to the original segments.

If matching segments aren’t found, the translator can create and add them to the TM database for future reference.

Benefits of Translation Memory

Professional translators rely heavily on translation memory for a number of reasons.

• Faster turnaround: Translation can be done faster when there are many matching segments in the translation memory.
• Higher earnings: With the aid of translation memory, it’s easier to translate more content, thereby earning more money.
• Lower costs: Translation memory allows reusing previously translated content to lower translation costs.
• Higher quality: Translation memory system can massively improve translation quality by reducing the room for human error.
• Consistent content: Translation memory allows ensuring consistency across all content through the use of the same translated segments.
• Improved collaboration: Centrally stored translation memory is easily accessible, which facilitates collaboration across teams.

Limitations of translation memory

Despite the advantages of translation memory systems, there are some drawbacks as well.

• Dependence on the previous translation: TM suggestions are only as good as the translation segments stored in its database.
• Limited application: While TM systems are great for repetitive or technical texts, they are less useful for creative, idiomatic, or unique texts such as literary works or marketing slogans.
• Time-consuming database creation: Although translation memory saves time, building and curating a TM database is a time-consuming process.
• Over-dependence on suggested translation: Relying heavily on TM suggestions can curtail the level of creativity and originality in the output.

What are translation memory use cases?

Given all the strengths and limitations of translation memory tools, what kind of applications are they suitable for? Generally, translation memories work best for texts that include a fair bit of repetition and jargon. The most common use cases include:

• Product documentation and manuals are not usually written from scratch for a new product or an update. Translation memory can speed up document translation and improve its consistency by reusing the language from existing materials.
• Self-service support materials are updated more frequently than product documentation, making them an even better use case for TM.
• Financial and legal texts usually have the same jargon across various sections and documents that can be reused by translation memory. Apart from speeding up the process, TM also improves consistency and accuracy, which is essential for financial and legal texts.
• Software and video game content is updated regularly, requiring the use of TM to achieve consistency between versions. This includes character names, their accessories, dialogues, and UI elements.
• E-commerce product descriptions contain similar phrases and descriptions across different products, thereby allowing TM to reuse them. Learn how to translate a website page here.

How to build a robust translation memory database

The quality of the translation output you deliver to your clients depends on how good the translation memory is. Let’s see how to ensure that your databases are up to the job.

Trust the language service provider, but verify

Before incorporating the segments translated by a new language service provider (LSP) into the main database, verify the quality of the matches. Thus, you will prevent the corruption of your TM with poorly translated content.

Don’t give your new language service provider (LSP) full access to your master translation memory. Instead, allow access in the read-only mode and create a working TM copy where they can commit their translation.

Automate linguistic quality assurance

Avoid running two translation memories in parallel by using the right QA tools.

Such tools automatically verify the quality of the translation and prevent errors from being saved into the translation memory. This is not limited to minor inaccuracies like spelling or punctuation mistakes but also covers more serious segment-level errors.

Conduct manual quality assurance

Even the most advanced automation tools won’t substitute an expert in linguistic quality assurance. An experienced and well-qualified linguist can pick up errors overlooked by automatic QA.

Saving the newly translated content in the database only once they’re verified will help you avoid quality issues not only in the current translation but also in future ones.

Provide necessary context

Context is essential for translators to provide accurate content. Always make sure your LSPs know where you will use their translations. This could be in the form of visuals or screenshots, or text fields with details like the domain, product family, or customer name.

When choosing a translation tool, make sure it allows you to share such contextual information.

Rank translation memory suggestions

If there are multiple translation memories involved in the process, some of them will naturally be of better quality than others. Before enabling access, give each of them a quality rating so that suggestions from trusted TMs appear at the top.

Update your TM

If not all your translations are done with the TM tool you’re using, make sure to update your database retrospectively.

For example, you may make some changes directly in the documents that you’re translating. To ensure future consistency, include all these changes in the database.

Constantly communicate with your customer and the translators

Whenever issues with the quality of the translation emerge, make sure to let your LSP know about them. In this way, you will be able to avoid the same mistakes in future translations.

The communication lines should be open both ways. Encourage the linguists that work for you to keep you up to date with any mistakes they come across.

This also applies to all alterations introduced by customers to communicate their preferred terminology and style.

5 Translation memory best practices

Once you put the effort into building your TM, follow these best practices to properly maintain it.

Recycle and reuse

Make use of previously translated content, such as manuals or product descriptions, whenever possible. By doing this, you'll boost exact and fuzzy matches and make the most of your translation memory tools.

Promote a centralized approach

To ensure consistency between translators and editors, provide them with the same translation software and TM databases.

Invest in translation memory maintenance

To ensure the quality of your translations, schedule regular cleaning sessions. Assign the job to an editor with expertise in the field and a good eye for detail.

Use standardized naming

Standardized naming will help you organize your translation memories.

It's a good practice to include information about the languages and the sector (legal, commerce, etc). In addition, many translation agencies like to keep translation memories for different clients separate so that they reflect their preferences.

Back it up

Translation memories are some of the most important assets in a translator’s work, so make sure to back them up along with all the other content on a regular basis.