Discourse Element Identification Integrated with Attention

To address the challenge of insufficient contextual dependency understanding in discourse element recognition, we propose a recognition model that incorporates an attention mechanism. This model analyzes sentence structure and positional information, using Bidirectional Gated Recurrent Units (BiGRU) and an attention mechanism to enhance the understanding of contextual information, thereby capturing the relationships among sentence components more accurately and improving classification accuracy. Additionally, we introduce Convolutional Neural Networks (CNN) for capturing inter-sentence dependencies. Experimental results show that our model outperforms existing baseline models across multiple datasets, confirming the effectiveness and practicality of the proposed method.