TermBoy – the AI Agent for Terminology

All four workflow schemas contain the same element: the AI agent TermBoy, which builds and maintains the termbase. TermBoy is a separate TTN product, firmly integrated into TTN TMS as a terminology agent and described in detail at www.termboy.com.

Figure 1: TermBoy – AI agent for terminology and visibility in chatbots (www.termboy.com)

From word list to term page

A classic glossary consists of two columns: source term and translation. What the term means, where it is used and which synonyms are permitted remains open – the translator adopts it blind. TermBoy instead produces a complete term page for every term: definition, image, translations into all required languages, synonyms, part of speech and gender, standards, references to Wikipedia and reference passages where the term is actually used within the organization. From the frequency of use and the quality of the sources it also calculates a reliability code.

Four steps – fully automatic

  1. Analyze: TermBoy searches the client’s website and related online sources, its text archives and its translation memories and identifies the specialist terms used most frequently.
  2. Define: a complete term page is produced for each term, with definition, images, standards, reference passages, synonyms, pronunciation and gender.
  3. Approve: if reliability is below 80 %, the entry goes to the client’s responsible subject-matter expert for review – notification by e-mail, SMS or WhatsApp.
  4. Publish and connect: publication with semantic tags on the web, inclusion in the client website’s sitemap.xml and provision in SDL MultiTerm for the CAT tool.

Interaction with TTN TMS

With every new order, TermBoy scans the source text and compares the terms found with the client’s termbase. New terms are researched and added while the translation project is still being created. The terminology is then available in three places: in the translator’s CAT tool, where the matching entry appears automatically in the translation window; in the QA tool, which enforces the use of the approved terms; and in the machine pre-translation, whose prompt receives the relevant entries segment by segment. For the short avalanche bulletin published daily in winter by the SLF in Davos, that amounts to up to 80 terms per translation prompt.

Visibility in AI systems

Search behavior is shifting from lists of hits to direct answers from chatbots. Anyone absent from those answers is invisible to a growing share of the market. Language models are based on transformer technology and need context, not individual search terms. Published terminology supplies exactly that context: every term becomes an indexable page with definition, translations, synonyms, standards and references, and is referenced in the client website’s sitemap.xml. This increases the likelihood that systems such as ChatGPT, Claude, Gemini, Perplexity or the AI Overviews of Google Search will classify the content correctly and cite it as a source. This approach is known as GEO (Generative Engine Optimization); it complements existing search engine optimization rather than replacing it.

Who benefits

  1. Translators: the terms appear automatically in the CAT tool as soon as a segment contains a specialist term; the QA ensures that only defined terms are used.
  2. Employees: new staff quickly find their way around the vocabulary of the organization – with images, descriptions and links to catalogs and web pages.
  3. Education and training: the terms used in the company match those of the vocational school and the SERI curricula, at no additional effort.
  4. Associations and public authorities: TermBoy reconciles the terminology with the official federal databases – important for consultation procedures and for avoiding isolated solutions.
  5. Clients and the public: all terms are published with semantic tags; consistent, publicly accessible terminology becomes a competitive advantage.

Models and effort

TermBoy is available in two ways. Order-driven, the termbase grows with the translation orders: a term is added as soon as it actually occurs in a translation. Research, term page, approval workflow, publication and CAT integration are included in the line or word price of the translation; no separate project budget is required.

As a comprehensive solution, Termboy builds a terminology database independently of individual projects, using the client’s website, translation memory, and documents. At the EU and the Swiss Federal Administration, defining and recording a term is estimated to cost around €10. Termboy reduces this cost to approximately one two-hundredth of that amount, while automatically keeping the terminology database up to date.

Examples of existing termbases – automotive engineering, science and digital marketing – as well as answers to frequently asked questions can be found at www.termboy.com. A demonstration based on your own website can be requested there or by e-mail to info@termboy.com.

Terminology beyond the boundaries of the individual order

A termbase created for a particular type of text – a bulletin, a catalog, a series of standards – does not remain confined to it. Once a term has been recorded, defined and approved, it applies to the entire communication of the organization: to the translations in all languages as much as to the original texts, the website, the training material and the information given to clients and authorities.

The effort is incurred once, the benefit remains: a term that has been clarified once never has to be discussed again in any later translation – regardless of language, order or year. That is precisely the difference between a glossary compiled for one project and then forgotten, and maintained terminology that keeps growing with every order.