Home / TERMGUARD
Private Sector
Stage 1, Stage 2

TERMGUARD

Σημασία-ΑΙ Ι.Κ.Ε. (Simasia-AI P.C.)
‹ Back to Open Call Winners
TermGuard is a neurosymbolic terminology engine that makes generative AI write term definitions to the ISO 1087 standard and proves they comply. A language model drafts a definition, a rule engine checks it (genus, differentia, circularity, negation, conciseness) and returns precise feedback, and the model redrafts until it passes. It runs inside a full terminology workbench with IATE grounding, TBX export and terminology-aware translation. In Stage 2 we integrate a European Model Hub model as the default backbone, publish the validator as an open-source AI-Builder module, and run beta tests with language-service providers.

Background

Generative AI is now used to write and translate specialised content, but it produces definitions that are fluent and confident yet invalid: circular, missing the broader concept, or defined by negation. Term storage exists (IATE) and text generation exists (LLMs), but there is no tool that checks whether a definition is structurally valid and grounded in an authoritative source. Translators, terminologists, publishers and public bodies working across the EU’s 24 official languages carry the cost.

Proposed solution

TermGuard combines neural and symbolic AI. An LLM drafts a definition guided by the ISO 1087 evaluation grid; a deterministic rule engine with packs for 16 EU languages scores it and explains every failed rule; the model then redrafts against that feedback. A separate semantic check tests whether the definition can replace the term in a real sentence without changing its meaning.

The loop sits inside a terminology workbench: upload a corpus, extract term candidates, draft and validate definitions, extract concept relations, ground terms in IATE and EuroVoc, and export to TBX (ISO 30042). Validated terms then constrain machine translation, with an automatic check that the approved terminology was used.

The system is provider-agnostic: any LLM can be the backbone, including European models. In our benchmark across six models and 90 runs, structural validity rose from 56% at first draft to 98% after the loop, in about two iterations. It is deployed as a containerised service on EU cloud.

Objectives

  • Make a European Model Hub model the default backbone of TermGuard and publish its benchmark
  • Release the ISO 1087 validator as an open-source AI-Builder-Suite module with a reusable Definition-QA workflow
  • Add a semantic validation layer and structural genus extraction for more EU languages, measured against expert judgement
  • Complete a trustworthy-AI / EU AI Act self-assessment and list TermGuard on the AIoD Marketplace and Business Navigator
  • Beta-test with language-service providers and publishers, and prepare the Stage 3 deployment with an adopter

Σημασία-ΑΙ Ι.Κ.Ε. (Simasia-AI P.C.)

Greece
Founded in Greece in 2025, Simasia-AI P.C. builds EU-hosted neurosymbolic language technology pairing LLMs with deterministic, standards-based verification. Its proprietary products—TermGuard, SimasiaStudio, and grounded digital assistants—deliver ISO-compliant translation and editing, already serving paying healthcare clients with full IP ownership.
Stay informed on AI opportunities
Get updates on new solutions, market trends, and adoption strategies delivered to your inbox.
The DeployAI project has received funding from the Digital Europe Programme (DIGITAL) under grant agreement №. 101146490 - DIGITAL-2022-CLOUD-AI-B-03
Stay informed on AI opportunities
Get updates on new solutions, market trends, and adoption strategies delivered to your inbox.
The DeployAI project has received funding from the Digital Europe Programme (DIGITAL) under grant agreement №. 101146490 - DIGITAL-2022-CLOUD-AI-B-03