【摘要】 命名实体识别和关系抽取是自然语言处理领域的两个重要基本问题。联合抽取方法被提出用于解决传统解决管道抽取方法中存在的一些问题。为了充分融合头实体和句子的语义信息,同时解决可能存在的重叠三元组问题,论文提出了一种新的实体关系联合抽取方法,主要通过序列标注的方式抽取实体关系。该方法主要使用条件层归一化(Conditional Layer Normalization)进行信息融合。同时,该方法还赋予了待抽取的头实体和尾实体不同的语义编码。实验结果表明,该方法在使用预训练的BERT预处理编码器的情况下,在NYT和WebNLG数据集上有很好的表现。更多还原

【Abstract】 Named entity recognition and relation extraction are two important basic issues in the field of natural language processing. The joint extraction method is proposed to solve some problems existing in the traditional solution pipeline extraction method. In order to fully integrate the semantic information of the head entity and sentence,and to solve the problem of overlapping triples,this paper proposes a new entity relation joint extraction method,which mainly extracts entity relation through sequence labeling. Its main feature is to transform the relational triples extraction task into a cascaded binary sequence labeling task,and use the multi-information fusion structure Conditional Layer Normalization for information fusion. At the same time,this method also gives different semantic codes to the head entity and tail entity to be extracted. The experimental results show that this method has a good performance on NYT and WebNLG datasets when using the pre-trained BERT preprocessing encoder.更多还原