The Volkswagen Group serves customers worldwide and delivers a vast range of documents in over 60 languages. Up to two billion words and documents must be translated within the Volkswagen Group every year, and the volume is rising. New communication streams and new demands regarding translation turnaround times meant that existing translation systems needed to be enhanced and expanded. Volkswagen turned to Azure AI, which offers a faster, more precise, and more cost-effective solution, with more advanced learning capabilities.
“In the past, we had to get someone who could translate documents for us, and now we can quickly translate them with a machine. It makes communication across languages faster and more effective.”
Nikolas Meyer-Aun, Head of Quality and Supplier Management for Languages, Volkswagen Group
High-volume, high-quality text and document translation
Translating up to two billion words a year in more than 60 languages sounds like an incredible feat, but it’s standard routine for the Volkswagen Language Services team. From owner’s manuals to presentation decks to board of management documentation, the Volkswagen Group’s language experts are always immersed in a wide variety of text and document translation. Many of the documents, such as owners’ manuals, are legally binding, so the Volkswagen Group must ensure high-quality translation.
For example, when text appears on the infotainment systems built into cars, the terminology that appears on the screen must match the text in any corresponding apps, owner’s manuals, related web content, and in other communication channels. “This means the quality of our translation memories is extremely important,” says Dirk Zehnpfund, Head of Language Services at the Volkswagen Group. A translation memory is a database in which segments of text are stored in the source language alongside the corresponding translation.
“The Volkswagen Group translation memories comprise over 100 million brand-specific translation texts and are used by all agencies worldwide that work with us,” explains Nikolas Meyer-Aun, Head of Quality and Supplier Management for Languages at the Volkswagen Group. The contents of these databases have been perfected by the Volkswagen Group over the years and fine-tuned to the terminology of the automotive industry.
Hand in hand with the machine
For the last seven years, Volkswagen Language Services has recorded a substantial increase in translation volumes—a trend that Zehnpfund says has become more pronounced with the Volkswagen Group’s digital transformation push. Furthermore, clients are expecting results within ever shorter timeframes. “We’ve received our first requests for regular translations to be delivered within 24 hours. Naturally, the high quality of our translations must not suffer due to urgency and time pressure,” says Zehnpfund.
Currently, the portal handles more than 16,000 projects every day. However, since the underlying technology translates content in a rules-based manner, investment is needed each time a new language is added. With so many languages utilized, this approach is expensive. In addition, rules-based technology limits both the speed of translation and total number of words that can be stored in the source and target languages.
Building a custom, cloud-based platform
To address these issues, Meyer-Aun’s team searched for cloud-based alternatives, which promised not only greater scalability, but also more flexibility —thanks to the neural machine translation models available in the cloud. After an extensive comparison of various providers, the company chose Microsoft Azure.
“We want to build a technical foundation on Azure Cognitive Services—on a collection of pre-trained models for machine learning and translation API,” says Meyer-Aun. The Volkswagen Group is training the Azure API standard translation models using its own translation memories, incrementally making them fit for the automotive manufacturer’s special terminology.
The Azure-based system manages translations into English, French, Portuguese, Spanish, and Chinese. The prototype translation system was fed with translation memories, developed, and tested in just six weeks.
There are many advantages to the cloud solution. Unlike the previous portal, the cloud solution is accessible by smartphone—a practical advantage for Volkswagen Group employees. The cloud streamlines processes by making the updates to translation memories available through uploads through the web interface. Previously, a service provider was needed to handle this task.
Translating full documents and complex file types, fast
The first iteration of the translation system processed text translation. This has been expanded to include document translation. With this innovation, the Volkswagen Group can quickly translate documents in multiple file types, such as .xls, .tlf, XML, PowerPoint, and others. The system currently translates an average of 325 documents per day, ranging from 1-page files to 1-million-character books. Document translation also keeps track of which translated piece corresponds to which section of the original document and can reconstruct the layout and structure of the original document. “In the past, we had to get someone who could translate documents for us, and now we can quickly translate them with a machine. It makes communication across languages faster and more effective,” says Meyer-Aun.
As a global company, the Volkswagen Group needs to stay on top of the newest updates on automotive regulation and safety legislation in countries around the world. Document translation allows the Volkswagen Group to quickly interpret and understand new legislation to determine what’s relevant to its business. “When government departments send out legislation, it's an unbelievable number of documents,” says Meyer-Aun, “and our teams need to check to see what’s relevant.” With this translation technology, the Volkswagen Group can interpret and share critical information quickly and ensure safety compliance around the world.
Safeguarding compliance and confidentiality
A deciding factor for the Language Services team was that none of the data—translation memories, documents to be translated, and trained models—was allowed to leave the European Union (EU) for data protection reasons. Ensuring compliance with the EU’s strict data protection regulations made Azure a natural choice. Microsoft can guarantee this because of the regions and geographies defined in Azure, meaning that the Volkswagen Group decides in what region on earth the created translation models should be stored, in line with the relevant data protection requirements. Free or open-source translation solutions do not ensure data security and confidentiality; Microsoft’s solution does.
In addition, users must accept the terms and conditions of usage, which specify that personal data is not allowed to be translated by machine translation. Even confidential documents are protected by strict EU requirements. “The translator gives us much better data security and data protection, which is essential for compliance with EU data protection laws,” says Meyer-Aun.
Taking advantage of big data
The Volkswagen Group continues to expand the capabilities and application of its machine translation. “Translation through a neural network helps boost quality and speed, and advances how we work with big data,” says Meyer-Aun. For example, mechanics all over the world are already accessing and recording parts and vehicle information in the Volkswagen Group’s central and regional databases using their own devices. They do so in their primary language, often using the customer’s vernacular, sometimes with misspellings. The system helps them quickly find technical product information based on this vehicle information. Thanks to cloud connectivity, the descriptions of the problems can be automatically translated into a standardized language form, collected centrally, tagged, and then made available worldwide in the market-specific languages.
The basis for this, in addition to Azure translation functions, includes other Azure services such as Data Lake, Machine Learning, Databricks, Stream Analytics, SQL Database, and Microsoft Power BI.
Looking ahead, the Volkswagen Group is eager to explore custom translator capabilities and find new ways to employ machine translation to meet the needs of customers and employees. The translation services are currently connected to an estimated 10-20 different platforms throughout the Volkswagen Group, and the organization is looking to expand this to more systems, including Office 365. And as Microsoft releases new features, such as the ability to translate scanned PDFs, The Volkswagen Group will continue to apply machine translation to new areas of its business.
“As the cloud environment is unlimited in its scalability, we can also offer these services to other Volkswagen Group brands without impairing performance,” says Meyer-Aun. “We are keen on exploring more possibilities from Microsoft, and we’re looking forward to continued improvement and a bright future together.”
Find out more about Volkswagen on Twitter, Facebook, and LinkedIn.
“The translator gives us much better data security and data protection, which is essential for compliance with EU data protection laws.”
Nikolas Meyer-Aun, Head of Quality and Supplier Management for Languages, Volkswagen Group
Follow Microsoft