What is a Translation Memory and How Do You Develop One?
Imagine translating the same words and phrases over and over. It sounds tedious, right? Fortunately, professional translators no longer need to. Today, 88% of translators rely on at least one computer-aided translation (CAT) tool to manage their projects—most of which include a translation memory to store and update translations.
Translation memory plays such a critical role that you need to know how it fits into your localization strategy. While you don’t need to develop one yourself (your translation team will do that), there are steps you can take to ensure you get the most from this technology.
What is Translation Memory?
Simply put, a translation memory is a database that stores words, phrases, sentences, and paragraphs that have been translated. Any time someone translates a word or phrase, known as a string, the translation memory automatically recognizes it and creates a new entry. Think of it as a bilingual dictionary that you curate over time.
How Does Translation Memory Work?
The translation memory matches the source text with translated text and adds both to the database as a single entity. This is called a unit. Once a unit has been created, the translation memory will flag similar segments as the translator works on the content.
Matches can either be 100% matches or fuzzy matches. A 100% match contains a segment identical to one from a previous project, while a fuzzy match is less than 100%. Although you can’t use fuzzy matches word-for-word, you can modify them for the new content.
The more content you translate, the less you need to translate each time manually. We’ve found that introducing translation memory enables 25% content re-use, significantly improving output ROI and consistency.
For example, let’s say you want to translate an e-commerce website. The translation memory would contain 100% matches for repeated fields such as size and color. If you’ve been translating product descriptions for clothing, the translation memory may automatically translate words such as “t-shirt,” “small,” “women’s,” and “red.” It may also suggest fuzzy matches for phrases used in similar product descriptions, such as polo shirts. In this instance, translation memory would save time and money and help ensure uniform brand language across the website.
Even if your content isn’t repetitive, fuzzy matches can be helpful—especially when the language pairs have a similar word order. That’s because segments with a similar sentence structure can serve as a starting point for a translation. And as you create more content, you’ll likely use some repetition in syntax to reinforce the tone of voice.
The Benefits of Translation Memory
Yet translation memory provides several benefits beyond eliminating the need to retranslate the same strings repeatedly. These include:
Consistent translation. Because it can automatically translate 100% matches, a translation memory ensures the language is the same through the content.
Cost savings. Translators can reuse strings without any cost (previously translated words and phrases don’t add to the word count).
Better quality. Linguists can refine translation memory to match a brand’s voice and tone.
Centralized database. Any department from any location can access translation memory. Team members can even process and transfer translation memories as translation memory exchange (TMX) files or Excel files.
Faster time to market. Translators can focus on original content, while the translation memory populates the previously translated strings. This reduces turnaround time—especially for larger projects.
Ultimately, translation memory benefits both translators and the companies that hire them as it increases efficiency and accuracy, all while reducing costs.
Translation Memory versus Machine Translation
However, it’s important to remember that translation memory is not the same as machine translation. Although translation memory automates the translation of predefined strings, it’s built by people who have translated your previous projects. Machine translation (MT), on the other hand, uses neural networks to interpret the source text and generate the translation. Both tools are helpful but aren’t appropriate for every type of project.
Translation memory is best for projects that require professional linguists with subject matter expertise. Translators may use it for tasks such as:
Marketing and advertising
Medical texts
Scientific research
Legal documents
This is because machine translation hasn’t reached the degree of linguistic nuance that is often necessary for adapting these types of content. Hiring human translators who use tools such as translation memory is usually the best solution for projects like these.
Machine translation is best for large volumes of text that don’t require as high of a degree of accuracy, such as:
Assembly instructions
Technical manuals
Customer support material
User-generated content
Although professional translators can prepare these types of content, machine translation is faster and more affordable. Case in point, Vistatec’s machine translation solution can increase turnaround by up to 40% and reduce costs by up to 50%. This method allows companies to allocate more of their budget to translating and localizing brand-sensitive content.
Translation Memory, Style Guides, and Glossaries
Now let’s look at two similar tools translators use: style guides and glossaries. Like translation memory, both of these help translators maintain a consistent voice throughout the content. However, style guides and glossaries serve separate roles in the translation process. For one, they’re typically developed before a translator begins work.
Below is a brief overview of each:
Style Guides
A translation style guide outlines a company’s rules about writing style. It’s typically written in the source language to explain the source elements and translation expectations. For example, the style guide would be in English if you want to translate an English-language website into Spanish. Linguists will then use it as a reference to ensure the translation matches the style and tone of the original content.
A style guide should also include the following:
Reference materials for the project
Descriptions of the target audience
Brand voice and tone
Spelling and punctuation preferences
Visual and formatting guidelines (e.g., logos, brand colors, and fonts)
Brand names and products
Acronyms
If you haven’t created one already, it’s essential to do so before you start a localization project. That’s because your style guide will play a vital role in the following:
Saving time by confirming your preferences.
Ensuring consistency across all localized materials.
Serving as a building block for a glossary and translation memory.
Glossaries
A translation glossary is a company’s list of standardized terminology that appears throughout the source text. It may also include company-approved translations for those terms and/or their definitions. Most glossaries have the following elements:
Terms in the source language
Definitions of terms
Approved translations of terms
What shouldn’t be translated (branded-related terms, product names, etc.)
Your language service provider can work with you to develop a glossary at the start of a project. This will speed up translation times by reducing ambiguity around word choice and ensuring all terminology matches the brand’s style and tone in the target language.
Translation Memory
Although your style guide and glossary are developed separately, they will help translators build a translation memory for your content. Instead of relying solely on dictionary definitions, translators can populate the translation memory with your preferred brand language. This further improves consistency across your multilingual content.
Final Thoughts
As you can see, translation memory is just one of the tools translators use to improve workflows and outcomes. And while it plays a key role, it can’t stand alone. Translators and clients must collaborate to curate this ongoing dictionary for each project.