In the realm of legal research and document drafting, accurate citation recommendations are vital for ensuring the relevance and credibility of legal arguments. This paper presents an innovative approach to enhance legal citation recommendations by leveraging clustered judgment networks and contextual BiLSTM models. We construct a comprehensive citation network from a substantial legal dataset, where nodes represent judgments and edges denote citations. The network is then segmented into meaningful clusters using Louvian clustering algorithm, enabling the capture of intricate relationships within legal texts. For each identified cluster, we train a BiLSTM model tailored to understand and predict context-specific citation patterns. This model effectively harnesses the sequential and contextual nature of legal documents, improving the accuracy of citation recommendations. When a query draft is introduced, its relevant cluster is identified based on its contextual embedding, and citations are recommended using the BiLSTM model of the corresponding cluster. Our method demonstrates significant improvements in the relevance and precision of citation recommendations, as evidenced by comprehensive evaluations against traditional citation recommendation systems. By integrating clustering and contextual modeling, our approach not only enhances the efficiency of the recommendation process but also ensures that the suggested citations are contextually pertinent and legally sound. This research contributes to the advancement of intelligent legal information systems, facilitating more effective and informed legal research and writing.