Creating glossaries for artificial intelligence

More accurate AI results start with glossaries that systematically map your specialist terminology.

oneword creates glossaries for use in production directly within machine translation systems and AI systems.

The purpose of a glossary for AI systems

A terminology database contains far more information than an AI system can process. While entries containing definitions, context sentences and grammatical information are valuable for human terminology work, they are generally unsuitable for integration into machine translation or a large language model in this form.

A glossary suitable for artificial intelligence requires careful selection and preparation:

This includes technical terms, product names, and brand-critical terms, for which, based on experience, AI systems tend to produce inconsistent or incorrect translation output

The mapping is carried out for each language direction, with all terms in the source language being assigned a preferred equivalent in the target language

The scope and file format depend on the technical requirements of the relevant machine translation system or AI system

A glossary that is too extensive can impair system performance – quality beats quantity

Would you like to know which terms in your corpus are relevant for a glossary?

Arrange a consultation now

What oneword does

Analysis and selection of relevant terms

We analyse your terminology, identify the terms that have the greatest impact on translation quality, and flag terms relevant for the glossary. The result is a prioritised selection – not an unfiltered list.

Glossary creation in all language directions

We produce glossaries in all required language directions, tailored to the requirements of your machine translation system or AI system. Any missing equivalents are specifically researched and added. Terminology will be provided in the file format required by your systems and processes.

Measuring effectiveness and updating

Once the glossary has been integrated, we will provide you with key performance indicators on its effectiveness and update the glossary as required, ensuring that your terminology remains effective even as your product portfolio grows.

Here’s how we do it

Introduction and overview

We’ll get an overview of your terminology resources, the AI systems or machine translation systems you use, and the language directions required.

Analysis and prioritisation

We determine the proportion of your database that is relevant to the glossary. You will gain a clear understanding of which terms are required and where foreign-language equivalents are missing.

Clean-up

Glossaries require clear guidelines. We highlight where this consistency is lacking in your data and help you with a targeted clean-up.

Glossary creation

The glossaries are produced using a semi-automated process in all the required language directions. As part of this process, potential stumbling blocks such as ambiguities and abbreviations are taken into account and addressed thoroughly.

Delivery in the appropriate file format

You will receive glossaries that are ready for direct integration into your machine translation system or AI system, or for passing on to your translation service provider.

Ensuring effectiveness

Once the integration is complete, we provide key performance indicators and update the inventory on an ongoing basis.

Our services at a glance.

When glossary creation makes sense

Glossary creation is the right course of action when AI systems or machine translation are in use, but the specialised terminology in the machine translation output is incorrect or inconsistent, or the post-editing workload remains consistently high. It is also worth taking this step even before new systems are introduced: A well-organised glossary can be put to productive use right from the start and will grow alongside your product portfolio.

uebersetzungsdienstleister-oneword-0788b

FAQs

What is the difference between a terminology database and an artificial intelligence glossary?

A terminology database is designed with people in mind: It contains definitions, usage notes, example sentences and grammatical information. An AI glossary is a condensed, format-specific version of this – selected and adapted for use in machine translation systems and AI systems.

Which AI systems and machine translation systems are supported?

We operate independently of any particular provider and adapt glossaries for common machine translation systems and AI systems. We will clarify exactly which file formats and interfaces are relevant during our initial discussion.

What happens once the glossary has been integrated?

The machine translation systems and AI systems use the glossary guidelines to determine how to handle technical terms. Your terminology is therefore applied consistently with the help of a glossary. We measure the effectiveness of the glossary and provide you with specific key performance indicators. For new products or projects, we update the corpus on an ongoing basis upon request.

Does a bilingual glossary always work in both language directions?

Glossaries are specific to a particular language direction and cannot simply be reversed. This is because all terms in the source language – whether preferred, permitted or prohibited – are assigned only to the preferred term in the target language. However, thanks to well-organised databases and clear usage information, glossaries can be generated at the touch of a button – and are therefore quickly available in any language direction.