基于位置辅助标记的实体关系抽取方法

为了解决在抽取过程中出现的关系三元组重叠问题，提出了一种基于位置辅助标记的实体关系联合抽取模型，使用BERT作为预训练语言模型，并且通过位置辅助矩阵方法，将关系三元组抽取转换成实体和关系的匹配问题，实现实体和关系的联合抽取，在中文数据集DuIE上进行了相关实验。实验结果表明，该模型抽取效果较好，提出的基于位置的辅助标记方法有效解决了关系重叠问题。

An Entity Relationship Extraction Method Based on Location-Aided Marking

In view of a solution of relation triplet overlapping in the extraction process, a joint extraction model of entity relation based on location-aided marking has thus been proposed. Taking BERT as the pre-trained language model, by means of position-aided matrix, the relation triplet extraction is transformed into the matching of entity and relation, thus realizing the joint extraction of entity and relation, with relevant experiments to be carried out on the Chinese data-set DuIE. The experimental results show that the model exhibits an improved performance in extraction, showing that the proposed location-aided marking method can effectively solve the problem of overlapping.