Neural Embeddings for the Elicitation of Jurisprudence Principles: The Case of Arabic Legal Texts

In the domain of law and legal systems, jurisprudence principles (JPs) are considered major sources legislative reasoning by scholars. Generally accepted JPs often used to support for a given case (JC). Although eliciting associated with specific JC is central task reasoning, it complex requires expertise, knowledge domain, significant lengthy human exertion This study aimed leverage advances in language modeling JP elicitation. We investigated neural embeddings—specifically, doc2vec architectures—as representation model elicitation using Arabic texts. Four experiments were conducted evaluate three different architectures document embedding models task. addition, we explored an approach that integrates task-oriented word embeddings (ToWE) (paragraph vectors). The results showed promising approach. paragraph vector distributed bag-of-words (PV-DBOW) architecture produced best this To how well ToWE performed task, graded relevance ranking measure, discounted cumulative gain (DCG), was used. achieved good normalized DCG 0.9 majority JPs. findings have implications understanding texts can be modeled semantics elicited embeddings.