Language varieties and dialects remain an open challenge for machine translation (MT). Plus, existing benchmarks rarely combine the evaluation of specific varieties with domain-specific knowledge. We introduce a termbase-derived MT benchmark for legal terminology across four German varieties: Germany, Austria, Switzerland, and South Tyrol (a low-resource German variety spoken in Italy). In the inter-variety setting, models translate Italian source sentences by selecting the correct South Tyrolean German term among semantically equivalent terms from other German varieties. In the intra-variety settings, we test sense disambiguation and ontological relation capabilities using homographic and ontologically related distractors for all four language varieties. We find that distinguishing language varieties remains difficult. High performance on low-resourced South Tyrolean terminology is partly driven by indirect exposure to terms shared with better-resourced varieties; when this overlap is removed, even the best-performing models collapse toward small models. We also find mild evidence of dominant-variety interference, although stronger systems more often err toward the geographically and culturally closer Austrian variety. Generally, locale-oriented adaptation appears more challenging than linguistic and semantic reasoning, where the ontological relation task is largely saturated and no generalized bias is shown against a specific variety. We conclude that model scaling mainly reinforces dominant language varieties and that LLM-based MT shifts part of the performance bottleneck from individual term rarity to the underrepresentation of domain-variety intersections across training and alignment stages.

The integration of artificial intelligence (AI) with terminology management (TM) has opened new avenues for enhancing efficiency and precision in both fields, necessitating standardized approaches to ensure interoperability and ethical application. The newly formed ISO/TC 37/SC 3/WG 6 represents the first dedicated initiative to study the standardization of the mutual improvements of AI and TM. This group aims to develop standardized frameworks and guidelines that optimize the interaction between AI technologies and terminology resources, benefiting professionals, systems, and practices in both domains. This article presents the state-of-the-art in the mutual relationship between AI and TM, highlighting opportunities for bidirectional advancements. It also addresses limitations and challenges from a standardization perspective. By tackling these issues, ISO/TC 37/SC 3/WG 6 seeks to establish principles that ensure scalability, precision, and ethical considerations, shaping future standards to support global communication and knowledge exchange.

We investigate the effect of terminology injection for terminology-constrained translation in a low-resource language variety, with a particular focus on off-the-shelf instruction-tuned Large Language Models (LLMs). We compare a total of 9 models: 4 instruction-tuned LLMs from the Tower and EuroLLM suites, which have been specifically trained for translation-related tasks; 2 generic open-weight LLMs (LLaMA-8B and Mistral-7B); 3 Neural Machine Translation (NMT) systems (an adapted version of MarianMT and ModernMT with and without the glossary function). To this end, we release LegISTyr, a manually curated test set of 2,000 Italian sentences from the legal domain, paired with source Italian terms and target terms in the South Tyrolean standard variety of German. We select only real-world sources and design constraints on length, syntactic clarity, and referential coherence to ensure high quality. LegISTyr includes a homonym subset, which challenges systems on the selection of the correct homonym where sense disambiguation is deducible from the context. Results show that while generic LLMs achieve the highest raw term insertion rates (approximately 64%), translation-specialized LLMs deliver superior fluency (∆ COMET up to 0.04), reduce incorrect homonym selection by half, and generate more controllable output. We posit that models trained on translation-related data are better able to focus on source-side information, producing more coherent translations.