In an interconnected world, businesses are grappling with the challenge of maintaining consistency and accuracy in translations across multiple platforms. DeepL, a key player in the machine translation (MT) space, has introduced a new solution to this problem: an AI-powered glossary generator. This tool allows users to generate glossaries automatically based on previously translated files, a step that previously required manual effort.

The glossary generator builds on DeepL’s existing glossary feature, which enables users to establish translation rules for words and phrases to maintain consistency across a team’s work. With this latest enhancement, users can upload files in various formats — such as .docx, .pdf, or .tmx — and DeepL’s tool can extract and compile a glossary, making translation customization quicker and more efficient. Supported language pairs include English to and from German, Spanish, French, Italian, Russian, and Japanese.

Christopher Osborne, DeepL’s Vice President of Product, emphasized the potential impact of this feature. “AI-powered translations are essential for businesses looking to overcome language barriers in today’s increasingly connected world, and DeepL’s powerful glossary tool takes this a step further by ensuring translations are personalized to a company’s unique phrases and needs,” Osborne told MultiLingual magazine. He also remarked how the tool improves the DeepL experience by offering “more efficient, accessible, and user-friendly” capabilities, ensuring teams can achieve the customization they need without relying on costly manual processes.

The glossary generator is currently only available to users of DeepL’s Pro Advanced and Ultimate plans. The feature is restricted to DeepL’s web platform and excludes the desktop, mobile apps, and API integration for the time being.

The DeepL support website acknowledges the need for human oversight in the translation process, stating, “Using files translated only via MT might lead to unsatisfactory results when generating a glossary. We recommend using files that were translated or reviewed by a human editor.”