Translation memory

Translating software application using translation memory

As the market of software translation and localization is growing rapidly, and project life spans becoming increasingly short, the need for constant documentation updates is ever rising, bringing to life new tools and solutions so as to enhance the human translation effort.

To simplify the translation of similar text content, specified translation memory (TM) database can be used. Implementing translation memory solution can improve productivity levels in the translation of, say, software applications text resources at a rate from 10% to 50%, depending on the text content, its consistency, and the localization/ translation software used. Translation memory is designed to improve the quality and efficiency of the human translation process, but it does not entirely replace human involvement. TM systems help the translator teams reach three major objectives: improve consistency of the translation (which is most likely to achieve), minimize turnaround time and, finally, with some deliberate analysis and planning, reduce the translation cost.

Translation memory stores the translated matching source and target language segments for future reuse. The TM systems basically consist of a database in which each source unit of a translation (usually equal to a sentence, though translation memory may contain translations of whole paragraphs), is aligned with the target unit. Typically, TM systems are integrated in translation/ localization software tools. Translation memory looks up the equivalent for each segment in the source text and automatically translates it into the target language. Besides, match values for searched elements are calculated.

Exact matching means the text segments in a revised source text match the original source text exactly. If the matching elements in the source text and translation memory are similar, but not identical, it is called fuzzy matching. Fuzzy matching occurs if the text segments in a revised source text are aligned with the translation of similar text segments from a previously stored translation memory. Duplicated exact matches are also often treated as fuzzy matches. Fuzzy matching function can be set to different levels of sensitivity based on particular criteria.
Sentences with match values below the margin set for fuzzy matching have to be translated from scratch. New and changed translation proposals will then be stored in the database for future use.
In conjunction with terminology database, TM system allows the translator to obtain a suggested translation for separate words (and word combinations), and if necessary, offer alternatives or point out different meanings of the word. Thus the translator does not need to use a separate electronic dictionary.

Most translation memories also contain some attributes, or additional information, such as the creation date, the user name, the project ID and the area of translation (for instance, legal, technical, etc.).

Although the translation memory systems offer outstanding advantages to the translator, accelerating the translation process, it needs considerable revision and validation of the results. For example, if terminology changes prior to the project update, the certain changes need to be inserted in the translation memory either. Besides, automatically leveraging on the previous translation using exact matches requires validation of the translated segments, or otherwise it can generate incorrect format strings, inconsistent translation, etc. Some of localization software suites, supporting translation memory, like Lingobit Localizer, include Validation function, which automatically verifies the translation and highlights the error-containing segments.