In large-scale systems engineering, Natural Language (NL) is the preferred medium for documenting requirements, because it facilitates communication between stakeholders with diverse technical backgrounds. To manage the ambiguity of NL, industry standards like the International Council on Systems Engineering’s (INCOSE) Guide to Writing Requirements (GtWR) establish formal frameworks to ensure that requirements are expressed with sufficient quality in order to support the downstream development processes. Moreover, the GtWR mandates the use of a glossary of domain-specific or project-specific terms to be referred by the requirements to ensure a consistent and clear use of the vocabulary. However, maintaining these glossaries manually in large-scale industrial projects is labor-intensive, error-prone, and difficult to scale, often resulting in a failure to consistently apply the relative INCOSE rules during the Elicitation and Analysis phases of Requirement Engineering (RE).?? To address these challenges, this thesis proposes a Retrieval Augmented Generation (RAG) solution that automates the consultation of project glossaries, as well as the automatic replacement of the original terminology when potentially ambiguous..