Automated Genre Classification of Urdu Poetry Using CNN-BiLSTM: A Deep Learning Perspective

The automated classification of Urdu poetry into distinct genres presents a significant challenge due to the language's rich morphology, contextual depth, and poetic constructs. This study proposes a CNN-BiLSTM-based hybrid deep learning model for genre classification of Urdu poetry, leveraging convolutional layers for local feature extraction and bidirectional LSTMs for context-aware sequence learning. The model is trained on a carefully curated dataset of diverse poetic forms and evaluated against state-of-the-art deep learning and traditional machine learning classifiers. Experimental results demonstrate that the proposed CNN-BiLSTM model achieves an accuracy of 94.2%, significantly outperforming conventional approaches such as SVM, RF, and standalone deep learning models like CNN and LSTM. Comparative analysis highlights the model's superiority in capturing long-range dependencies and semantic coherence, making it highly effective for nuanced genre differentiation. These findings establish a robust framework for Urdu poetry classification and pave the way for future advancements in computational literary analysis.