Sweden-based language service provider (LSP) Semantix won a EUR 34m (USD 41m) contract with the European Union in December 2020.
The tender covers translation services for the European Commission’s Directorate-General (DG) for Internal Market, Industry, Entrepreneurship and SMEs, which aims to promote economic growth and an “entrepreneurial culture” in Europe. Specifically, the translation services will help implement the Single Market Transparency Directive (SMTD).
To comply with this EU directive, member states must notify the European Commission, other member states, and stakeholders of new proposed regulations related to goods and “information society services.”
Member states use the Technical Regulations Information System (TRIS) to transmit relevant notifications, which typically consist of a title, summary, short explanatory notice, and text of the draft technical regulation. There is a three-month “standstill” period after each notification is issued, which allows the EC and other member states to issue reactions, comments, and opinions before the notifying state adopts the draft.
The current contract with the incumbent provider, Luxembourg-based LSP AMPLEXOR (acquired by Paris-based rival Acolad in 2020), ends February 28, 2021. The new contract will last up to 48 months, including three automatic 12-month renewals.
Semantix will fulfill the contract in concert with ESTeam, the software company headed by Trados creator Jochen Hummel. Poland-based LSP Summa Linguae Technologies will also contribute linguistic expertise.
Kaleidoscope of Language Pairs
According to the call for tenders, “in principle, all notifications and notified texts are translated into all official languages of the EU (taking into account any upcoming requirements concerning Gaelic) and then made available to the Member States and the stakeholders.”
In practice, notifications and supporting information and documents can be received in any of the EU’s 23 official languages and will be translated into some or all of the official languages, though the exact number of target languages depends on the nature and length of the text. Translation from and into Irish is not required until at least January 2022.
Non-EU countries that participate in the notification procedure (namely Iceland, Liechtenstein, Norway, Switzerland, and Turkey) are required to submit their drafts, notification messages, and documents in one of the languages of the EU; translation from and into Icelandic, Norwegian, and Turkish is not provided.
For a sense of scale, the call for tenders offers statistics on the numbers of source language pages translated in 2018.
For draft technical regulations and impact assessments, 21,260 pages were translated into 142,670 target language pages. German, with 111 drafts totaling 3,826 pages, was by far the most voluminous source language for draft technical regulations. English, with 19,294 target language pages, was the most requested target language in this category.
English was also the most common source language for messages exchanged during the procedure, with 373 messages totaling 885 pages (out of a total of 3,180 source language pages). The most requested target languages for these messages were French and German, with 2,543 and 2,614 pages translated, respectively.
In addition to these two categories, an estimated 1,402 pages of procedure-related documents requiring ad hoc translation, as well as 14,041 pages of final text, were also translated.
Semantix’s bid for the contract was evaluated according to five criteria: establishing a sustainable team of appropriate translators; applying specific techniques during the translation process to ensure reliability, accuracy, and terminological consistency; employing quality assurance and control procedures to produce ready-to-use translations; assigning texts for translation to translators and revisers; and organizing management of the contract.
In particular, Semantix was lauded for its quality control procedures, which feature domain experts participating throughout the translation process “via the collaboration features of the workflow system.”
“For complex texts, on top of the regular workflow, the tender envisages an additional first step of manual editorial control by a source language native speaker domain expert to identify key concepts and expressions and to reduce the number of possible queries during translation,” the award letter stated. “The solutions, comments and references are directly visible and filterable for linguists in the computer-assisted translation (CAT) tool.”
Britta Aagaard, Semantix’s Chief Business Officer, elaborated on the company’s daily capacity monitoring of resources, which is based on a multilingual knowledge system (MKS) called Coreon. Unlike terminology bases, she said, information in an MKS can easily be processed by machines.
“For us, it is the basis for domain classification, content recycling, terminology recognition and verification, and semantic quality assessment,” Aagaard told Slator. “We use it to match competences and capacity at all times, to drive automated quality procedures, and to optimize the whole process of translation.”
To cater to all language combinations, specializations, and volumes, Aagaard estimates approximately 2,000 linguists will be involved in work related to the contract at some point.
“The challenges are mainly due to the large number of possible language combinations (522 in total) where not all combinations are equally blessed with sufficient quality data and linguistic assets nor highly specialized linguists for all possible domains,” Aagaard added. “In these cases, relay processes can be applied, though obviously requiring additional quality assurance steps while observing timely deliveries.”