Chinese Entity Relation Classification via Fusing Granularity Information and Gated Recurrent Mechanism
Zicheng Zhong · 2021
Entity relation classification is a foundational natural language processing task, which plays an important role in text analysis. At present, most of the methods for Chinese entity relation classification have the problem of segmentation errors, which will cause the accumulation and propagation of errors in the subsequent steps. For this problem, we propose a deep learning framework that integrates word-lever information and character-lever information, which utilizes the efficient fusion of multi-granularity features and gated recurrent mechanism (FGGRM) to dynamically learn semantic information with less effect of segmentation errors. Through the comparison of multiple methods on two datasets, our model shows excellent performance and ability to learn features.