电梯安全事故领域命名实体识别研究

Research on Named Entity Recognition of Elevator Safety Accident Domain

知识图谱技术是解决数据多源异构的有效解决方法,目前在很多领域得到了应用,而命名实体识别(NER)是自动构建领域知识图谱的关键步骤,但在电梯安全事故领域尚未见有命名实体识别(NER)的相关研究。论文针对构建电梯安全事故领域知识图谱的应用目的,提出基于针对中文文本分词改进的BERT预训练模型与BiLSTM-CRF相组合的模型实现对领域非结构化文本中的实体进行自动抽取,提出了适合电梯安全事故领域的命名实体识别(NER)模型。论文收集整理了500余份电梯安全事故文本作为实验语料数据集。通过实验表明,相较于传统命名实体识别模型,论文所使用的模型识别效果有显著的提升。

Knowledge map technology is an effective solution to solve the problem of multi-source heterogeneous data.It is ap-plied in many fields at present,and named entity recognition is a key step to automatically build the domain knowledge map.Howev-er,there is no related research on named entity recognition in the field of elevator safety accidents.Aiming at the application pur-pose of building the knowledge map of elevator safety accident field,this paper proposes a model based on the combination of BERT pre training model improved for Chinese text segmentation and BiLSTM-CRF to automatically extract entities from unstructured text in the field,and proposes a named entity recognition model suitable for elevator safety accident field.This paper collects and col-lates more than 500 elevator safety accident texts as the experimental corpus data set.Experiments show that compared with the tra-ditional named entity recognition model,the recognition effect of the model used in this paper is significantly improved.