Translation memory technology is designed to assist the translators in ‘remembering’ what they, or translators before them, translated and how they translated it.The ‘memory’ part is a database that stores source – target pairs of text segments.  Translators use TM tools to store these pairs as they create translations. They can later view, search for, and reuse these translated segments. The advantages of using this technology are several:

• Consistency in translation within and between documents, products, projects
• Collaboration between teams of translators who have access to each other memories (or work with a centralized, server-based memory)
• Learning and reference tool for the translators new to the project or product
• Faster turnaround and lower cost for related translation projects, as well as upgrades and updates to the previously translated content (when compared to translating that content from scratch)
• Better translation quality (although some might disagree with this statement)

These benefits can be obtained through use of TM matches.

When translators use a TM tool, the content being translated is automatically compared to what was translated in the past and is stored in the translation memory. If any matches are found during this comparison, the tool offers translators existing translation along with the original source text and assigns a score to that match expressed as a percentage. If the original source segment is the same as the current segment, then it is an exact, or 100%, match. If the value is less than 100%, then it is a fuzzy match. 

If the context in which the 100% match occurs is the same, translators can automatically reuse it. If the context is different, translators might need to check and, if necessary, modify the translation to be appropriate for the new context. Nowadays, TM tools offer translators the flexibility to distinguish between exact matches that occur within the same context (in-context exact matches, aka ICE matches) and exact matches that occur in a different context. In-context matches can be safely reused without any further input from the translator.

Repetitions are 100%-match ‘wannabes.’ Once the first occurrence of the repeated segment has been translated, the remaining occurrences will come up as 100% matches.

Translating repetitions and 100% matches takes significantly less effort than translating a no-match segment or even a fuzzy match. For this reason, translation services providers (companies and freelance translators) typically offer a significant discount for translation 100% matches and repetitions and some discount for translation of fuzzy matches.

Fuzzy matches require more effort from the translator, depending on the degree of matching. Translation services providers often base their pricing on different levels of matches. For example, they might offer 50% or similar discount for fuzzy matches within a certain range and charge regular rate for any matches below lower limit of the set range.  Thus, if the lower limit of the range is a 75% match, a standard rate will apply to any matches of 74% or less, as for the translation of new text, because it takes the translator as much or almost as much time to translate a segment of 74% match as a non-matching segment.

There have been many animated discussions between the buyers and sellers of translation services about offering discounts for repetitions, exact and fuzzy matches since the TM technology gained wide acceptance. For better or for worse, it is now a standard practice for sellers of translation to offer discounts for TM matches in projects involving text types that lend themselves well to repetition.

Although still meager, empirical research on effects of TMs is starting to appear. For example, Eye-tracking and Translation Memory Matches by Sharon O’Brien (Perspectives: Studies in Translatology. Volume 14, Issue 3, 2007; PP. 185-205) suggests that translators cognitive load when translating fuzzy matches is lower than for no matches and even lower when translating exact matches. As more translation scholars focus on the convergence of translation technology with the task of translation, we might eventually have some scientific data that will lead us to revisit the debate on matches, discounts and quality when working with a TM.