Philipp Ursprung is responsible for project management, translation technology and terminology at the Language & Translation Services department in Credit Suisse.
In his role as business project manager, he took an active part in designing and integrating a Machine Translation system into Credit Suisse and its translation workflows.
Philipp Ursprung studied translation at the Zurich University of Applied Sciences (ZHAW) and holds a degree in translation and translation technology from the University of Surrey (UK). He worked in different positions both for language services providers and corporate language services departments, where he focused on project management and the introduction and optimization of CAT tools and translation processes and workflows.
The Introduction of Machine Translation at Credit Suisse
Recent technological developments have boosted the use of Machine Translation (MT). Although MT might never replace human translators, online MT tools and smartphone applications have quickly become part of everybody’s lives. At the same time, there is an increasing number of successful use cases of MT with human post-editing in corporate translation scenarios.
For an institution that handles sensitive data, such as a bank, the popular online MT tools are a potential source of risk for data confidentiality reasons, despite the advantages they offer to a global company and its employees. Credit Suisse (CS) addresses the need to protect confidential data by introducing its own secure MT system to offer all staff a secure alternative to online MT tools for their translations for comprehension purposes. Credit Suisse Machine Translation (CSMT) is based on the statistical machine translation (SMT) system Moses and offers MT engines customized with CS and banking specific corpora in 22 language combinations.
This approach results in high-quality MT engines that are not only available to all staff via the intranet but also seamlessly integrated into the production workflow of the CS Language & Translation Services department at the pre-processing stage, as an addition to the existing leverage of translation memory and terminology database. Since the rollout of CSMT in early 2015, translation projects in 16 language combinations contain MT suggestions that allow for efficiency gains with both external translation providers and in-house language specialists. The CS approach is to offer MT suggestions for every translation in language combinations with MT engines available despite the large variety in content, document type and purpose of the translation requests.
In addition to the technical implementation of the MT system, the focus was on offering training the internal language specialists at CS in light of their shifting from translators to post-editors who never come across an empty target segment anymore. Being early adopters of MT at least in the financial industry, onboarding the existing external translation services providers early enough in the process was pivotal.
The integration of the MT system required a great deal of time and money but proved to be feasible and worthwhile, both with regard to security and financials. In a global company like CS, MT will be a must for the future to handle the increase in content requiring translation with limited translation budgets. While current SMT systems have their limitations, they have become easy to implement into translation workflows and CAT tools, helping the post-editor’s learning curve and the acceptance of the new technology.
Future efforts around MT in CS include designing and applying engine retraining cycles, systematic quality evaluations to close the gaps to fully leverage the technology and improve user-friendliness of the MT application website.