基于深度学习的远程监督关系抽取方法研究

Research on remote monitoring relationship extraction method based on deep learning

为更好判定远程监督语句中的实体语义关系,实现对语言处理信息的准确提取,提出基于深度学习的远程监督关系抽取方法。借助远程监督方法,获取关系三元组中已存储的信息参量,再通过待学习数据标注的方式,完成远程监督关系的抽取数据集构建。在此基础上,设计监督执行框架,利用已定义的句子级别特征条件,实现对待抽取标签的学习处理,完成基于深度学习的远程监督关系抽取方法研究。实验结果表明,所提方法可同时调度的远程监督语句数值量更大,而所需的分辨等待时间却相对更短,可在有效判定远程监督语句中实体语义关系的同时,实现对语言处理信息的准确提取与分配。

In order to better determine the entity semantic relationship in the remote monitoring statement and realize the accurate extraction of language processing information,a remote monitoring relationship extraction method based on deep learning is proposed. With the help of the remote supervision method,the information parameters stored in the relationship triad are obtained,and then the extraction data set construction of the remote supervision relationship is completed by the way of data annotation to be learned. On this basis,the supervision implementation framework is designed,the sentence level feature conditions defined are used to realize the learning processing of extraction tags,and the remote supervision relationship extraction method based on deep learning is completed. The experimental results show that the proposed method can schedule more remote monitoring statements simultaneously,but the required waiting time is relatively short. It can effectively determine the semantic relationship of entities in remote supervision statements,and realize the accurate extraction and distribution of language processing information.