Chinese named entity recognition based on dual-channel dilated convolution neural network
Chaoyi Wan, Ruiqin Wang, Qishun Ji, Yimin Huang · 2024
Chinese Named Entity Recognition (CNER) is a crucial information extraction task. However, the current mainstream models encounter issues regarding practical application in large-scale datasets or real-time scenarios. Firstly, there is a deficiency in the real-time processing of text streams due to insufficient speed and efficiency of the models. Secondly, data quality significantly affects model performance. To address these challenges, this paper proposes a novel Named Entity Recognition model (BiLSTM-DDCNN-AAT-CRF) that integrates a dual-channel dilated convolution and attention mechanism. The improved version of dilated convolutions is incorporated to enhance speed while ensuring recognition accuracy. Additionally, the IOBES tagging scheme is employed to mitigate potential model performance degradation resulting from inaccurate or inconsistent annotated data. Experimental results on the Sogou News dataset and a general dataset demonstrate a significant improvement in named entity recognition compared to existing models. Furthermore, ablation experiments validate the effectiveness of the attention mechanism and dual-channel approach proposed in this paper.