Chinese relation extraction based on lattice network improved with BERT model
Zhengchun Zhang, Qingsong Yu · 2020
Relation classification is a basic and important task in the field of natural language processing(NLP). There are already many researches on English dataset, but the researches on Chinese dataset are very few. Due to the particularity of Chinese language, most existing methods suffer from the two main problems of segmentation error and polysemy. To sum up, the problem of segmentation error can be solved fairly well by many models, take lattice model for example, which can segment Chinese word precisely. But the problem of polysemy has not received enough attention. In this paper, we take advantage of BERT model to deal with the problem of polysemy. The experimental results show that our model achieves good result and outperforms baseline model.