W-core Transformer Model for Chinese Word Segmentation

Hai Lin, Lina Yang, Patrick Shen-Pei Wang

Abstract: Chinese word segmentation is an important research content in the field of Natural Language Processing (NLP). In this paper, we combine the Transformer model to propose the Window Core (W-core) Transformer for the tasks. In this model, W-core can preprocess sentence information according to the characteristics of Chinese and incorporate features extracted by the Transformer model. Experimental results show that the W-core Transformer model can improve the effect of the original Transformer model on Chinese word segmentation. Finally, we improve the performance of W-core Transformer by increasing the number of encoder layers and oversampling.