TransPerfect’s Private Neural Neural Machine Translation Portal

TransPerfect’s Private Neural Machine Translation Portal Diego Bartolomé, José Masa TransPerfect Passeig de Gràcia 11, Esc. B 5è 2a 08007 Barcelona, Spain {dbartolome,jmasa}@translations.com Abstract We will present our solution to replace the usage of publicly available machine translation (MT) services in companies where privacy and confidentiality are key. Our MT portal can translate across a variety of languages using neural machine translation, and supports an extensive number of file types. Corporations are using it to enable multilingual communication everywhere. 1 Introduction Machine translation (MT) is widespread today1. Companies are using it extensively both for productivity increase and thus turnaround time and cost reduction, and also for gisting or understandability in many situation such as e- discovery. At TransPerfect, we have developed a neural machine translation platform that can be installed on premises or on our own cloud to guarantee data confidentiality and control, link client-specific neural MT engines to it, and enable supervised and unsupervised learning2. 2 Access to the platform The access is through a URL (to be presented at the conference), and can be customized for each  client. Our main features are: Single Sign On: no need for specific usernames or passwords, users at our clients can access with their company e-mail and password.  IP address range restriction: only users accessing through a pre-defined range of IP addresses are allowed into the system. This is © 2018 The authors. This article is licensed under a Crea- tive Commons 3.0 licence, no derivative works, attribution, CC-BY-ND. essential for security in our top clients like banks or pharma companies.  Real-time translation of plain text and documents: users can translate plain text and also more than 40 file types, including scanned PDFs and Office documents.  Neural MT engines: neural MT engines are available in more than 25 languages, with a supervised and unsupervised learning option. Supervised means that the engines learn from linguists’ feedback, and unsupervised refers to self-learning capabilities. A functionality to suggest a better translation is available, as well as automated language detection.  Reporting: powerful reporting is available to enable real-time tracking of number of processed words, quality of the engines, and other business KPIs.  Data storage: we delete data after 24 hours, and some clients have even more restrictive policies to delete translated plain text immediately and documents after they are downloaded. 3 Additional features Besides the above, we are currently integrating additional features that have been commonly requested such as customization of glossaries and do not translate lists, seamless integration with our human post-editing services, and addition of speech-to-text and text-to-speech as input and output modes, respectively. References [1] TransPerfect, The Year of Artificial Intelligence in Translation, http://www.transperfect.com/blog/the- year-of-AI-translation. [2] TransPerfect, MT: Robot Intelligence Technology with a Human Touch, TransPerfect blog, http://www.transperfect.com/blog/machine- P´erez-Ortiz, S´anchez-Mart´ınez, Espl`a-Gomis, Popovi´c, Rico, Martins, Van den Bogaert, Forcada (eds.) Proceedings of the 21st Annual Conference of the European Association for Machine Translation, p. 341 Alacant, Spain, May 2018.