Digital Palimpsests Leveraging Large Language Models for the Decipherment of Transnational Socio-Legal Archives

Abstract

The paper examines how a large language model (LLM) can be used to uncover the socio-legal texts of different countries, including international treaties and human rights laws, as well as international legal agreements. The model was tested on the main performance indicators: accuracy (92%), precision (90%), recall (89%), and F1-score (89.5%). These findings demonstrate the high success of the LLM in the analysis of complicated legal language in different legal systems and languages. The LLM was evaluated against the conventional manual systems, which normally take 5-7 days to interpret, and the accuracy is 85%. On the contrary, only 2-3 hours were required by the LLM to process the same documents, and the accuracy was considerably high, 92%. This shows the model's high efficiency and consistency. The model was also tested on the performance in various forms of socio-legal texts. It was found to be accurate 94% in cross-border legal agreements, 93% in human regulations, and 91% in international treaties. It means the model can interpret many legal documents, including those with multilingual components and complex legal terminology. The LLM's ability to reveal suppressed meanings, ambiguity, and more precise interpretations of these texts makes it a useful resource for legal researchers, policymakers, and international organizations. The results indicate that the LLM is a better tool than traditional practices in terms of speed, accuracy, and comprehensiveness, and thus can be used in future research in the area of socio-legal studies. The use of publicly available datasets, however, has certain limitations, such as the fact that it may not reflect all legal frameworks. The next-generation research may increase the number of cases and narrow the model to specific fields of law.