This paper presents a Conv1D-based approach for multi-class classification of legal citation texts, aiming to improve the accuracy and efficiency of automated legal text classification. The proposed method leverages Convolutional Neural Networks (Conv1D) with max-pooling and softmax activation to capture hierarchical features in sequential data, enabling effective multi-class categorization of legal citations. The model addresses limitations in previous studies by incorporating advanced neural network architectures and optimizing for the nuanced structure of legal texts. The approach is evaluated on a dataset of Australian legal cases from the Federal Court of Australia, with results showing significant improvements in weighted F1-score compared to traditional methods such as Random Forest, SVM, and MLP. The Conv1D model demonstrated a weighted F1-score of 0.57, indicating its effectiveness in balancing precision and recall while considering class distribution. The study contributes to the ongoing development of NLP applications in the legal field, offering a robust and efficient solution for automated legal text analysis. Future research directions include exploring additional neural network architectures, incorporating domain-specific embeddings, and extending the model to accommodate evolving legal language trends. The integration of Large Language Models (LLMs) is also discussed as a promising avenue for enhancing the capabilities of legal text classification systems.This paper presents a Conv1D-based approach for multi-class classification of legal citation texts, aiming to improve the accuracy and efficiency of automated legal text classification. The proposed method leverages Convolutional Neural Networks (Conv1D) with max-pooling and softmax activation to capture hierarchical features in sequential data, enabling effective multi-class categorization of legal citations. The model addresses limitations in previous studies by incorporating advanced neural network architectures and optimizing for the nuanced structure of legal texts. The approach is evaluated on a dataset of Australian legal cases from the Federal Court of Australia, with results showing significant improvements in weighted F1-score compared to traditional methods such as Random Forest, SVM, and MLP. The Conv1D model demonstrated a weighted F1-score of 0.57, indicating its effectiveness in balancing precision and recall while considering class distribution. The study contributes to the ongoing development of NLP applications in the legal field, offering a robust and efficient solution for automated legal text analysis. Future research directions include exploring additional neural network architectures, incorporating domain-specific embeddings, and extending the model to accommodate evolving legal language trends. The integration of Large Language Models (LLMs) is also discussed as a promising avenue for enhancing the capabilities of legal text classification systems.