Custom neural machine translation engines helps Seimas translators to significantly boost their productivity and efficiency

March 23, 2020

Client

The Seimas of the Republic of Lithuania is the main legislative body of Lithuania. Its main tasks include enacting laws and amendments to the Constitution, adopting the budget, confirming the Prime Minister and the Government, and supervising their activities.

Challenge

As a single-chamber legislator, the Seimas works with extremely large amounts of information and documents and must effectively communicate and cooperate with countless international organizations on daily basis. Due to highly sensitive content, translations are done in-house, which requires a lot of time. Considering this challenge, there was a pressing need for a solution that would ensure fast and smooth multilingual information exchange, provide accurate translations, improve efficiency of internal translators, and allow them to process more translations in a shorter period of time.

Solution

The Seimas knew that high-quality machine translation was the answer. However, they needed custom translation engines that were adapted to their specific needs. By relying on vast multilingual in-house data combined with translation memories provided by the Seimas, Tilde developed custom machine translation engines that were adapted to the legal domain, and produced high-quality translations for English-Lithuanian, Lithuanian-English, Russian-Lithuanian, and Lithuanian-Russian.

Results

Thanks to the custom machine translation engines that were seamlessly integrated into SDL Trados, translators of the Seimas have improved their translation efficiency and are able to process larger quantities of translations in less time, while maintaining the required quality and the integrity of terminology and style. The use of private machine translation engines also ensures Seimas with full data security and confidentiality as translation data is neither stored nor analysed, and gets deleted immediately after the translation is received.