Translation systems

Glossa Group offers a comprehensive portfolio of services in the fields of translation, editing and technical support. This includes in particular the development and enhancement of translation systems.

Machine translation (MT) systems have undergone continuous development since the 70’s. Their use requires in-depth know-how so that the final result can be used.

Translation memory systems that store translated data in a form that allows it to be automatically retrieved during future translations are equally widespread.

Glossa Group is developing solutions to increase efficiency in future translation processes by mastering the challenge of integrating and networking translation memory systems and machine translation systems.

• Translation memory systems
• Machine translation

Translation memory systems

The main idea behind translation memory systems is their ability to recognise previously translated text segments and to automatically transfer these segments to texts that are currently being translated. The automatic process described reduces the amount of work required for large translation projects by up to 60% and at the same time leads to an increase in quality because it excludes several translation variants.

The key capability of these programs consists in supporting the human translator's memory with its memory and search functions. They store all translated files in a form that for each new translation permits automatic scanning of old translations in a matter of seconds, determines matches on the basis of a comparison with the source text and automatically embeds them in the target text (pretranslation).

This means that the translator does not need to search through hundreds or even thousands of files until he or she finds a suitable, corresponding passage for the current translation, and also that there is no risk of creating changes in recurring text passages and the probability that inconsistencies will occur is therefore reduced.

Machine translation

Machine translation (MT) has its roots in rule-based systems from the 60’s and 70’s. Today’s newer hybrid approaches are combining rule-based systems with statistics and examples of existing translation (Example Based Machine Translation, or EBMT).

Machine translation still requires post-editing work by human translators to achieve good results. In general, it is only suitable for gathering an initial impression about the contents of texts, for example in bidding files or simple e-mails, or for intelligence gathering (MT for information, MT for dissemination and MT for communication).

As seen by GLOSSA GROUP, there are two challenges to be solved as regards increasing productivity and the use of machine translation:

1. the automated evaluation of machine translation output (since human evaluation is relatively expensive and time consuming) and
2. the integration of translation memory technology in machine translation

GLOSSA GROUP has developed internal systems and has collaborated in recent years with former members of LISA in exchanging benchmarking data and creating trustworthy automated metrics and a framework for an improved automated evaluation of MT output as well as standardising the exchange of data between MT and TM.

GLOSSA GROUP has gathered extensive knowledge in the field of MT and multi-language output of data with a short “shelf life” (its value decreases very rapidly), e.g. stock prices, weather, and other current information and Internet news services.