nEYron: Implementation and Deployment of an MT System for a Large Audit & Consulting Corporation

nEYron: Implementation and Deployment of an MT System for a Large Audit & Consulting Corporation Artur Nowakowski1,2, Krzysztof Jassem1,2, Maciej Lison1, Rafał Jaworski2, Tomasz Dwojak2 1 Poleng, Pozna´n, Poland {name.surname}@poleng.pl 2 Faculty of Mathematics and Computer Science, Adam Mickiewicz University, Pozna´n, Poland {name.surname}@amu.edu.pl Karolina Wiater, Olga Posesor EY Poland, Warsaw, Poland {name.surname}@pl.ey.com Abstract This paper reports on the implementa- tion and deployment of an MT system in the Polish branch of EY Global Lim- ited. The system supports standard CAT and MT functionalities such as translation memory fuzzy search, document transla- tion and post-editing, and meets less com- mon, customer-specific expectations. The deployment began in August 2018 with a Proof-of-Concept, and ended with the signing of the Final Version acceptance certificate in October 2021. We present the challenges that were faced during the de- ployment, particularly in relation to the se- curity check and installation processes in the production environment. 1 Business Need On March 6, 2018, the Polish parliament adopted a law that laid down rules for the Polish Agency of Audit Surveillance regarding the control of au- diting companies. The law states that “Documents presented by the audited company for the needs of the surveillance are drawn up in Polish or the au- dit company provides their translation into Polish.” The law forced auditing companies to provide Pol- ish translations for large volumes of English texts. That triggered the idea, at the Polish branch of EY Global Limited (EY Poland), that the cost of the task might be reduced if it were assisted by a trans- lation engine. EY Poland contacted the company Poleng Ltd. (Poleng) to verify the possibility of us- ing their product, TranslAide Workspace, for the © 2022 The authors. This article is licensed under a Creative Commons 3.0 licence, no derivative works, attribution, CC- BY-ND. task. During initial discussions, EY Poland came to the conclusion that it might be beneficial for the company to have the software installed and run- ning on site. 2 The Story of the Deployment 2.1 TranslAide Workspace The first phase of the deployment began in August 2018. The deployed system was based on TranslAide Workspace, which combined computer-aided translation (translation memory with fuzzy search and segment-by-segment edit- ing) with a generic machine translation engine, not trained specifically on the in-domain data. The task consisted in replacing the existing translation engine with a new one, dedicated to the customer. The deployment was divided into the Proof-of- Concept (POC) and Final Version stages. The POC machine was to be installed in the Linux environ- ment to make the initial deployment easier for the Poleng team. There were no explicit expectations regarding the quality of the translation imposed on the POC version. However, moving forward to the Final Version stage was conditional on acceptance of the POC by the customer – including transla- tion quality, which would be checked by human specialists from the EY corporation. The Final Version – all of the system components, includ- ing model training – was expected …

Content truncated to 500 words