Construction of Chinese named entity recognition model based on TIMP text information feature extraction

Abstract

To improve the efficiency of Chinese named entity recognition, this study optimizes existing models to address the insufficient context feature capture and difficulty in recognizing polysemous words when processing complex text information. Given the grid long short-term memory network model, text information memory perception module, text information adaptive fusion module, and conditional random field are added to enhance the model’s ability to capture contextual information and feature fusion effect. Finally, an improved lattice long short-term memory network model is designed for Chinese named entity recognition tasks. The results indicated that the new model had higher recognition accuracy than the conventional model in benchmark performance testing, with the highest recognition accuracy reaching 0.98 in both datasets. In practical applications, the model achieved recognition accuracy of over 95% in 10 different types of Chinese named entity recognition tasks, with the highest reaching 99.15%. In addition, the average recognition time of this optimized model was as low as 0.06 seconds, far less than the other three compared models. Therefore, the designed model can provide a more efficient and accurate technical means for Chinese named entity recognition tasks.