A translation memory (TM) is a database that shops "segments", which could be sentences, paragraphs or sentence-like units (headings, titles or components in an inventory) that have beforehand been translated, so as to aid human translators. The translation memory shops the source text and its corresponding translation in language pairs referred to as "translation units". Particular person phrases are dealt with by terminology bases and usually are not inside the domain of TM. Software programs that use translation reminiscences are typically known as translation memory managers (TMM) or translation memory techniques (TM programs, to not be confused with a translation administration system (TMS), which is another sort of software program centered on managing the strategy of translation). Translation memories are typically used in conjunction with a devoted laptop-assisted translation (CAT) device, phrase processing program, terminology administration methods, multilingual dictionary, and even raw machine translation output. Research signifies that many companies producing multilingual documentation are using translation memory programs. In a survey of language professionals in 2006, 82.5% out of 874 replies confirmed using a TM.

Utilization of TM correlated with textual content kind characterised by technical terms and easy sentence structure (technical, to a lesser diploma advertising and marketing and monetary), computing skills, and repetitiveness of content material. This system breaks the source text (the textual content to be translated) into segments, looks for matches between segments and the supply half of previously translated supply-goal pairs saved in a translation memory, and presents such matching pairs as translation full and partial matches. The translator can settle for a match, replace it with a recent translation, or modify it to match the supply. In the final two circumstances, the new or modified translation goes into the database. Some translation memory methods search for 100% matches solely, i.e. they'll only retrieve segments of textual content that match entries within the database precisely, while others make use of fuzzy matching algorithms to retrieve similar segments, that are introduced to the translator with differences flagged. Typical translation memory systems only seek for text within the supply segment. The pliability and robustness of the matching algorithm largely determine the efficiency of the translation memory, although for some applications the recall price of actual matches could be excessive sufficient to justify the 100%-match approach.

Segments where no match is found will have to be translated by the translator manually. These newly translated segments are saved within the database where they can be used for future translations in addition to repetitions of that phase in the current text. Translation memories work greatest on texts that are extremely repetitive, equivalent to technical manuals. They're also useful for translating incremental adjustments in a beforehand translated document, corresponding, for example, to minor changes in a brand new model of a person guide. Historically, translation memories haven't been thought-about acceptable for Memory Wave Protocol literary or inventive texts, for the simple purpose that there's so little repetition within the language used. Nonetheless, others discover them of value even for non-repetitive texts, as a result of the database sources created have value for concordance searches to find out applicable utilization of terms, for high quality assurance (no empty segments), and the simplification of the overview course of (supply and target phase are at all times displayed together whereas translators must work with two documents in a conventional overview setting).

Translation memory managers are most suitable for translating technical documentation and paperwork containing specialized vocabularies. Making certain that the translated paperwork are constant, together with common definitions, phrasings and terminology. That is important when totally different translators are engaged on a single undertaking. Enabling translators to translate documents in a large variety of formats without having to own the software program sometimes required to course of these codecs. Accelerating the overall translation course of; since translation recollections "remember" beforehand translated materials, translators should translate it only as soon as. Lowering prices of lengthy-term translation projects; for instance the text of manuals, warning messages or series of documents must be translated only as soon as and can be used a number of occasions. For big documentation projects, savings (in time or money) thanks to the use of a TM bundle might already be apparent even for the first translation of a new venture, but normally such financial savings are only obvious when translating subsequent variations of a challenge that was translated earlier than utilizing translation memory.

The idea of "translation reminiscences" is based on the premise that sentences used in previous translations may be "recycled". However, a guiding principle of translation is that the translator should translate the message of the text, and not its component sentences. Translation Memory Wave Protocol managers don't simply match into present translation or localization processes. So as to reap the benefits of TM technology, the translation processes have to be redesigned. Translation memory managers do not presently support all documentation codecs, and filters might not exist to support all file sorts. There's a studying curve associated with utilizing translation memory managers, and the programs should be customized for best effectiveness. In instances the place all or a part of the translation process is outsourced or handled by freelance translators working off-site, the off-site workers require special tools to be able to work with the texts generated by the translation memory supervisor. Full variations of many translation memory managers can price from US$500 to US$2,500 per seat, which can signify a considerable investment (although decrease value packages are additionally obtainable).