This poster presents an ongoing project aimed at developing a bilingual (French–Spanish) lexical infrastructure for contemporary architecture, designed according to TEI and LMF standards and enhanced through AI-assisted modelling techniques. The resource is conceived not only as a digital lexicon but as an open infrastructure that supports public engagement, scholarly reuse, and cross-disciplinary interaction between architecture, linguistics, and digital humanities.

The project is grounded in two parallel web corpora of architectural discourse (2000–2025), collected from magazines, blogs, institutional publications, and professional platforms. Term extraction was performed using NLP-assisted workflows, after which key lexical units were manually curated and structured following a TEI–LMF hybrid model (Romary 2003; Francopoulo et al. 2006). The lexicon integrates hierarchical, synonymic and bilingual equivalence relations, enriched through contextual embeddings (CamemBERT and BETO) and guided paraphrasing techniques. AI models were employed to identify semantic proximities, propose candidate relations, and assist in drafting definitional paraphrases, while final decisions remain expert-guided.

The poster demonstrates three layers of innovation:

1. Standard-based modelling of a specialised lexicon
   The TEI–LMF implementation allows the resource to function as an interoperable lexical infrastructure, facilitating sustainability, export to linked open data formats (e.g., SKOS), and potential integration into platforms such as the PanLatin Lexicon (ILC-CNR). The encoding captures conceptual relations, contextual examples, morphological data, and bilingual mappings, offering a rich representation of architectural terminology for researchers, translators, and educators.

2. AI-supported enrichment workflows
   The project evaluates how AI can assist with corpus exploration, lexical clustering, semantic alignment, and paraphrase generation without replacing expert control. The workflow demonstrates how distributional embeddings help identify conceptual networks within architectural discourse, contributing to a dynamic understanding of emerging terms (e.g., ecological metaphors, resilient-city frameworks). This component will be illustrated through prototype visualisations and small demos derived from the current database.

3. Public engagement and participatory potential
   Beyond scholarly applications, the lexicon is designed as a public-facing infrastructure. Architectural terminology carries strong societal relevance—urban planning, sustainability, environmental metaphors—thus offering opportunities for interaction with students, architects, cultural institutions, and local communities. Future development includes participatory annotation modules, public contributions on conceptual interpretations, and integration into teaching activities in translation, architecture, and digital humanities. This aligns the resource with DARIAH’s vision of infrastructures that enable openness, shared creativity, and socially responsive scholarship.

The poster will present the current beta version of the lexicon interface, examples of TEI–LMF entries, AI-assisted enrichment results, and visualisations of the conceptual networks extracted from the corpus. By illustrating both methodology and public engagement potential, this contribution showcases how a specialised digital lexicon can evolve into a hybrid infrastructure—technically robust, socially meaningful, and adaptable to transnational collaborations in digital humanities.