Context-Aware Query Interpretation in Legal Information Systems

Abstract

The need for user-specific context, the diversity of jurisdictions, and the complexity of legal language all contribute to the delay in accessing accurate legal information. Conventional keyword-based retrieval and even semantic search engines frequently return irrelevant and/or incomplete results due to their inability to incorporate those subtleties. To enhance the retrieval of legal information, this study proposes a system known as Context-Aware Query Interpretation (CAQI). The CAQI system has three main components: user profiling, query reformulation, and ontology-driven semantic enrichment. The model uses legal ontologies to resolve polysemy and hierarchical relationships. To refine query interpretation, context vectors encode user role, jurisdiction, and task intent. The method generates relevance scores that align with textual meaning and contextual appropriateness, utilizing a hybrid ranking function that combines lexical similarity (BM25) with semantic embeddings. The evaluation was validated by a user survey with legal professionals and performed on a curated dataset of 500 annotated legal queries. Outcomes demonstrate substantial improvements in retrieval accuracy, with CAQI outperforming baseline keyword and semantic models in Precision@10, Mean Reciprocal Rank (MRR), Discounted Cumulative Gain (DCG), and Contextual Relevance Score (CRS). To improve decision support, legal research efficiency, and user trust in intelligent legal information systems, these findings suggest that CAQI can "think like a lawyer" by dynamically aligning queries with jurisdictional and task-specific contexts.