Abstract: Legal language is inherently prone to indeterminacy, presenting significant challenges for drafting, interpretation, and translation. Ambiguity, polysemy, and context-dependent meanings can undermine precision in statutes, contracts, and judicial decisions, leading to divergent interpretations and practical uncertainties. Traditional approaches in legal linguistics and doctrinal analysis, while providing qualitative insights, are limited in their ability to systematically quantify and model semantic vagueness across large corpora of legal texts. This study proposes a computational lexico-semantic model designed to analyse indeterminacy in legal terminology. Leveraging recent advances in natural language processing (NLP), including transformer-based language models (GPT), contextual embeddings, and ontology-driven semantic frameworks, the model provides a probabilistic assessment of term vagueness. A multi-domain corpus comprising constitutional, criminal, and financial law texts was constructed to evaluate the approach. Semantic similarity measures, context-sensitive embeddings, and GPT-derived probability scores were applied to identify and quantify degrees of lexical indeterminacy. Preliminary results demonstrate that the model effectively detects terms with high semantic ambiguity, revealing patterns of indeterminacy that align with expert legal annotations. Visualisations of semantic probability distributions highlight the domains and contexts in which vagueness is most pronounced, offering insights into term usage, frequency, and contextual dependency. The findings have significant implications for legal drafting, automated translation, and AI-assisted legal reasoning. By providing a quantitative, scalable framework for analysing legal indeterminacy, the proposed model contributes to improving the precision, consistency, and interpretability of legal texts, while supporting future research on real-time monitoring of evolving legal terminology.

Keywords: Legal indeterminacy, lexico-semantic modelling, NLP in law, GPT language models, semantic vagueness, ontology-based legal analysis, computational legal linguistics

Download: | DOI: 10.17148/IMRJR.2025.021201