Schmalz succeeds with Contentserv and AI-based translation

As a global market leader, the company faces a major challenge in providing relevant information about its diverse products available online. With customers speaking a wide variety of languages, creating different country-specific websites is crucial. To manage product information efficiently and centrally, Schmalz has been leveraging Contentserv‘s powerful Product Information Management system (PIM) at its headquarters in Glatten in the Black Forest, Germany.

Previously, Schmalz’s translation process for its vast amount of product information was very time-consuming and relying on external translation agencies came with considerable costs. The aim was to optimize translation processes, making them faster and more cost-efficient. Schmalz wanted to reduce its dependence on external service providers and at the same time increase translation quality. Marco Ade from Product and Content Marketing at Schmalz reports that the translation agency was initially involved as an additional safety net in case the STS and DeepL process did not run smoothly. However, this concern was quickly dispelled.

The new program now processes translation requests via a REST interface. According to Sebastian Böhringer, the DevOps specialist at Schmalz who co-developed the program, the Schmalz Translation Service operates as a standard web service with a Java Spring Boot backend. The modern interface is based on Vue framework. The frontend communicates with the backend via HTTP or REST. In addition to the PIM system, other data sources like Schmalz's CMS and a knowledge transfer tool are also connected to the STS program.

STS users can select how the translation job is processed based on each language. For example, DeepL can handle specific translations but, sometimes, it‘s more beneficial to hire an external translation agency, or involve colleagues fluent in the local language. This is an option, especially if no translation glossary is available for a language in DeepL.

Schmalz's product and content marketing team populated its professional DeepL account with glossaries, translating specialized terminology into the required languages. This approach addresses challenges that a generalist translation AI like DeepL might struggle with. Schmalz also created glossary databases and stored them in DeepL for translations from German to English and from English to other languages, including Chinese, French, Italian, Japanese, Dutch, Polish, Russian, and Spanish.

Once DeepL receives the translation job for the required languages from STS via the API and completes it using AI, it compares the translation with the various language glossaries stored by Schmalz and makes necessary adjustments. The revised translation then flows back to the STS program.

At this point, the translation job is marked as "Approval". This means that the Deepl translation is distributed by STS to local colleagues with the appropriate access rights for review and approval. The STS provides reviewers with relevant context attributes to identify which Schmalz product the translation relates to. Additionally, the STS shows whether a local glossary was already available during DeepL‘s AI translation. For example, a Slovakian colleague can view the original text and the translation result within the STS interface and make any necessary edits directly.

If the translation result is satisfactory, it can simply be approved by clicking on "Approve". The translation service is then transferred back into the PIM system via the web service, removed from the job list, and archived. Schmalz gives colleagues one week to complete the approval process.

The results have convinced Schmalz across the board. Faster, better, and cost-effective translations are now achievable by combining Contentserv with the new Schmalz Translation Service (STS). The specific outcomes are:

• Quality improvement: With the support of local language glossaries, Deepl delivers more consistent and reliable translations of product information from Contentserv compared to the translation agency. The results are superior due to their consistent and reliable quality.
• Increased output: Since the new translation solution has been available, Schmalz translates more comprehensively than before, with no fundamental limits.
• Time savings: Automation has allowed Schmalz to save 99% of the time previously spent on organizing translations.
• Cost reduction: Schmalz saves around €70,000 annually through AI-driven translation and the automated process integrated with the PIM. Previously, achieving the higher output would cost around €150,000 to €170,000, whereas the API connection of the STS solution with DeepL costs less than €100 per month.

Automating the process with the help of AI has paid off for Schmalz. The STS has demonstrated the value of AI-assisted automation and localization of product information and texts. This approach can also be applied to various scenarios, including those involving context attributes that describe different industries.

Marco Ade views this as just the beginning: With the ongoing development of AI applications and language models, it‘s only logical that text production will soon be scaled up. There are many opportunities for Contentserv to combine PIM and AI to create an even better product experience.