Multi-feature Chinese semantic role labeling combined with global information

Semantic role labeling serves as a crucial approach to obtaining semantic information and enabling shallow semantic analysis. Currently, the BiLSTM-CRF model is the primary method used for semantic role labeling. However, this model has many network parameters and is unable to effectively capture semantic information in long sentences. To address these issues, this paper proposes a CNN-BiLSTM-MaxPool-CRF fusion model for Chinese semantic role labeling. The model utilizes MaxPool to sample and extract the output of the BiLSTM network to optimize the network structure. Convolution kernels of differing sizes are utilized to capture local features of sentences, and these features are then combined through average pooling to form new feature vectors. These new features incorporate semantic information of the sentence context and are inputted into the model alongside multi-level linguistic feature groups such as part of speech and sentence phrase structure. Through multiple sets of experimental demonstrations, the method proposed in this paper has demonstrated significant improvements in the performance of the semantic role labeling model.