基于知识图谱与渐进学习的典型事故预案自动化生成方法

Method for Automatic Generation of Typical Accident Plans Basedon Knowledge Graph and Progressive Learning

为自动化生成典型事故的可用预案,提出了基于知识图谱与渐进学习的典型事故预案自动化生成方法。使用基于历史预案知识抽取的知识图谱构建方法,将历史事故预案文本信息作为数据源,在确定知识图谱模式层内容后,数据层使用基于BiLSTM-CRF模型的知识抽取,提取历史预案知识信息,经融合、更新处理后,构建为三元组式知识图谱。将需应急处理的事故文本信息、所构建知识图谱,输入基于渐进学习的事故预案自动化生成模型。利用融入渐进学习思想的半监督Markov随机游走模型,在知识图谱中,计算预案知识迁移概率,将正例迁移概率大于负例的预案知识作为匹配的预案措施,完成预案自动化生成。经测试,所提方法具备高效率预案自动化生成优势。

To automate the generation of available contingency plans for typical accidents,a knowledge graph and progressive learning based automated generation method for typical accident contingency plans was proposed.Using a knowledge graph construction method based on historical contingency plan knowledge extraction,the text information of historical accident contingency plans was used as the data source.After determining the content of the knowledge graph pattern layers,the data layers used knowledge extraction based on the BiLSTM-CRF model to extract historical contingency plan knowledge information.After fusion and update processing,the knowledge graph was constructed into a triplet type knowledge graph.Input the accident text information that needed emergency response and the constructed knowledge graph into the automatic generation model of accident contingency plans based on progressive learning.Using a semi supervised Markov random walk model incorporating progressive learning ideas,calculate the transfer probability of contingency knowledge in the knowledge graph,and use the contingency knowledge with a transfer probability greater than negative cases as matching contingency measures to complete the automatic generation of contingency plans.After testing,the proposed method has the advantage of efficient automated plan generation.