This is one of the 52 terms in “The Language of Content Strategy” published by XML Press in 2014 and the contributor for this term is Stefan Gentz.
What is it?
A repository that contains translated source and destination language pairs.
Why is it important?
Reduces translation time and cost by reusing translated content from the repository.
Why does a content strategist need to know this?
Professional human translators use specialized tools to translate content. These tools are called Computer Aided Translation tools (CATs). A CAT tool consists of a translation memory, a terminology database, and an editor to translate virtually any kind of content.
During translation, the source language content is segmented by the CAT tool into small logical units. Such a unit is usually a full sentence (or heading, list item, table cell, etc.) in your content. This unit is called a segment. These segments are translated into the target language by a machine translation engine, a human translator, or both. Together, a source and target segment form a translation unit (TU). These TUs are stored in a linguistic database called translation memory (TM).
Modern translation memory systems enrich these database entries with additional metadata. Perhaps the most important is metadata that defines the semantic context. Other metadata types include content domain, date and time, translator, quality score, data source, and data type.
When new content is translated, the CAT tool analyzes each segment and tries to find a match in the TM. Matches can range from “no match” to “fuzzy matches” to 100%, or exact, matches. Modern CAT tools add context matches on top of that: “101% matches” that help eliminate false positives due to ambiguities or different context.
With every translation, the TM grows. The bigger the TM is, the more previous translations can be recycled. This dramatically speeds up translation and improves the quality and consistency of new content.