Optimization of translation memories

Reviewing terminology and consistency, checking for spelling mistakes or empty content: quality assurance is a basic component in the translation process. The checking functions integrated in a translation memory system can only be used for the documents to be translated, but not for the translation memories (TM) themselves that correspond with the translation.

As a result, other methods have to be implemented to check TM databases. Source texts that are unsuitable for translation and compromise the segmentation, changes made by the customer that are not adopted in the TM, inconsistencies through the assignment of multiple language service providers, possible terminology violations through changes in a terminology database, etc. – the list of potential sources of mistakes that can hinder the quality and consequently the effectiveness of the use of TM databases is long. Moreover, an ill-considered move to another translation memory system often leads to undesired match and value losses of a TM despite the interoperability and compatibility publicized by the system manufacturer. This is due to the TMS’ own “tag languages”, i.e. the different handling of program- and format-specific additional information within the translation segments.

KERN Global Language Services handled a project involving three TM case studies from which a range of measures were identified that show how TM databases can be checked, cleaned and thus optimized. The results are in:

Case study 1:

Avoiding terminology violations

The analyzed TM has been filled with content by KERN Global Language Services for many years and comprises a large number of translation units. In addition, the customer and the translators of KERN Global Language Services use a comprehensive terminology database, whose scope has increased over time and whose entries have been changed. The terminological consistency of the TM should therefore be checked.

Formal errors due to formats unsuitable for translation or segmentation of the source texts were also identified. In the process of TM optimization, the TM underwent de-formatting; tab stops, line breaks and superfluous spaces were removed; a standardization of punctuation marks, indications of measurements and abbreviations was performed; and numbering and bullets were checked for inconsistency. The TM was also checked for duplicates, inconsistencies and segmentation errors, and underwent spelling and terminology checks. All incorrect terminology was manually corrected. The result was a cleaned TM, which now offers the customer valuable matches in accordance with the database, and the quality was thus improved.

Case Study 2: TM migration

The customer collaborates with another language service provider in addition to KERN Global Language Services. The aim is the seamless merging of both language service providers’ TMs into a master TM, which can then be used by both service providers across projects. For cost reasons, the customer initially decided upon a casual migration in the past, which on the one hand – as expected – led to a loss of matches both when merging and when recognizing matches for subsequent translations, and on the other hand to inconsistencies. In order to prevent this, the already merged master TM was analyzed, freed of TMS- and format-specific additional information through a de-formatting process, and proofread with a formal checker for duplicates and inconsistencies in the source and target language. Furthermore, violations in the use of terminology were automatically detected and subsequently manually corrected on the basis of a newly prepared terminology database.

The TM optimization particularly improved the consistency of the source and target languages. Through the de-formatting, it was ensured that no segments were lost when importing the translation units into a new master TM.

Case Study 3: Combining the glossary and TM

A customer-specific but obsolete terminology database and an additional glossary in the form of an Excel document, which contains the complete and updated terminology portfolio of the customer, were available. The terminologists at KERN Global Language Services decided, in agreement with the customer, to delete the obsolete terminology database and instead edit the Excel document according to terminological aspects and then import it back into the terminology database. Because, for a certain period of time, it was not possible to use the complete and updated terminological specifications when creating translations, or rather because the specifications had only recently been developed, terminology inconsistencies occurred in the TM. In order to check the terminology, the TMs were analyzed with a formal checker and finally corrected.

The revised and optimized TM and the updated terminology database were fed back into the translation memory system, so that they were available to KERN Global Language Services’ translators from then on.

Quality through optimized TM databases

Despite the difficulty of checking TMs, KERN Global Language Services shows that steps to optimize a TM are altogether useful and even necessary to guarantee the quality of translations produced with the aid of translation memory systems in the long term. With the help of formal checkers and open source TM editors, every TM can be analyzed. Analysis and other reports, in which the potential errors of a TM are listed, subsequently allow for the evaluation of necessary optimization measures. In this context, it is also recommended to check whether source documents are suitable for the translation process and, where necessary, adapt them. Customer-specific editorial manuals or style guides help to create rules for writing in a manner suitable for translation as well as customer-specific style regulations. With the optimization of your TM and a sophisticated terminology management system, you can achieve long-term quality improvement. For these reasons and more, we encourage you to ask KERN Global Language Services for an optimization of your TM for the next translation order!