Attention-CNN with Multi-Task Learning for Chinese Named Entity Recognition

Abstract

Named entity recognition serves as a cornerstone in natural language processing and has garnered extensive research attention due to its significance in various downstream applications. Owing to the intricate nature of Chinese texts, characterized by complex syntactic structures and the lack of explicit word boundaries, conventional NER methodologies often encounter difficulties in simultaneously optimizing recognition accuracy and computational efficiency. To address this issue, the study proposes a named entity recognition algorithm that integrates attention mechanisms with Convolutional Neural Networks, incorporates into a Transformer-based bidirectional encoder framework for training. A multi-head self-attention mechanism is employed to capture the global semantic information of the text, and multi-task learning is introduced to construct the final model. When evaluated on datasets with sample sizes of 200, 1000, and 3000, the proposed model consistently outperforms the baseline models in terms of precision, recall, and F1 score. Specifically, under the low-resource setting with 200 samples, the model achieves a precision of 98.62%, a recall of 98.10%, and an F1 score of 98.36%. In terms of inference efficiency, the model processes at a speed of 2618 tokens per second. The experimental results indicate that this method can be widely applied in various fields such as information extraction and text understanding, providing strong technical support for related research.