Terminology integration: a neologism of the future
by Marguerite Meyer
Not only language specialists but also non-experts are familiar with the free translation tool DeepL with its column view showing source and target language. For some time now, however, users have had another function at their disposal: An on/off button at the top right with the note "glossary". Anyone who clicks on it can now specify how certain words are to be translated.
Example: Your English source text contains the word "experts". In the German translation the word "Experten" is used. However, if we want DeepL to translate "experts" into "Spezialisten" instead of "Experten", all we need to do is enter the desired terminology in the glossary function and DeepL will automatically adjust the translation.
Terminology enforcement: a neologism with a great potential
Among computational linguists and language technologists, the novel use of specialist terminology is called "terminology integration", or also "terminology enforcement", a term also used by the Swiss company TextShuttle. DeepL is thus by far not the only machine translation provider to make this technology available. One prominent example is TextShuttle, a spin-off company from the University of Zurich. TextShuttle developed a terminology enforcement function in both its customer-specific and generic MT engines.
What is special about this approach? Not only are individual expressions and terms saved – but also entire terminology databases can be stored and used together with the MT system. In other words, the engine is "forced" to use the uploaded terminology, hence the alternative term "terminology enforcement " (= "terminology integration").
Language Box tested this innovative function.
Our test run: terminology enforcement as a further milestone?
The procedure on TextShuttle's online platform is quite simple and similar to the glossary function of DeepL: The customer can independently enter the desired terminology in the corresponding field – and thus have the translation automatically adjusted.
Another possibility is to upload your own terminology database as a tbx file to control the machine accordingly – with just a few mouse clicks, and without needing any help from TextShuttle. A log file shows whether the tbx file was successfully uploaded.
Once the terminology database or the tbx file has been uploaded, the machine immediately inserts the correct technical terms within the CAT tool used – provided the customer works with a TextShuttle MT plug-in, such as those available for SDL Trados Studio, Wordbee and Across. According to TextShuttle, connection to other CAT tools is possible upon request. Thanks to the plug-in (i.e., the connector between the CAT tool and TextShuttle's MT system), translators use their MT engine directly in the CAT tool, including the uploaded terminology.
We consider the integration of the desired terminology via upload of the tbx file into the machine as well as the result before, during and after our test translation (whether online or within the CAT tool) to be highly satisfactory: definitely yet another successful milestone in the field of neural machine translation!
Fine-tuning and details: just a matter of time
Despite the noticeable increase in quality that terminology enforcement brings to machine-translated texts, however, post-editing is not eliminated. A machine is still a machine – and even with a clean and reliable terminology, declinations, plurals and other language-specific peculiarities sometimes need to be put through a final human polish. However, TextShuttle is optimistic in this respect. When asked by Language Box, for example, whether terminology enforcement works in every case, the software producer answers confidently:
"There will never be a one hundred percent guarantee with artificial intelligence. However, we have had very good experiences so far and are constantly improving the mechanism."
It will be interesting to see how terminology enforcement continues to develop. Language Box is staying on the ball and will continue to report on this in the future.