Translation Memory (TM)

Texts are often revised slightly for new product versions. When these revised versions are retranslated, it does not make sense to have someone retranslate those portions of the text that are unchanged. As a result, technology called translation memory (TM) is used to store texts and their translations, broken down into small pieces (usually sentences) called segments. 

When a new version of the text is translated, the TM tool processes the text and automatically replaces those parts that are unchanged with the previously created translations. This allows the human translator to focus on the new content. In the case of text that is only partially changed (so-called “fuzzy” matches), the translator is given the nearest matches and their translations for reference and editing. This process is called leveraging. Translation memory is most efficient for revisions of existing texts, but is sometimes used within projects where there is significant duplication between components (e.g., online help and user manuals) to create internal leveraging. The quality of results from translation memory depends on the quality of previous translations. The efficiency depends on how well the database is maintained and how proficient translators are in using the tool.