基于事件画像和案例推理的社区工单处置

Community Work-order Disposal Based on Event Portrait and Case-based Reasoning

近年来,随着政府数字化转型的不断深入,越来越多的12345政务热线工单下发到社区进行处置。工单文本信息通常较为稀疏,主题序列涵盖城市治理方方面面。社区管理人员对工单进行处置往往花费较长时间,无法满足群众实时响应的需求。为了提升社区工单处置的质量和时效性,该文提出了一种基于事件画像和案例推理的工单处置决策方法。首先,基于统一标准地址库以三元组方式构建地名地址基因库用以获取地名中的谱特征,构建树集合以表征地址基因之间的层次关系,利用地址基因之间的关联关系对缺失地址元素进行补全和还原;其次,为了充分发掘社区工单文本的局部特征和全局特征,该方法通过基于BiGRU、Self-Attention、CNN、CRF的组合神经网络对社区工单事件进行有效提取;最后,在构建社区事件历史案例库的基础上使用关键词提取并计算事件之间的相似度。对比实验结果表明,该方法相较于其他方法能够取得更好的性能。

In recent years,with the deepening of the government’s digital transformation,more and more 12345 government hotline work-orders are issued to the communities for disposal.Work-order text information is usually sparse,and the topic sequences cover all aspects of urban governance.It often takes a long time for community managers to handle these work-orders,which cannot meet the needs of the people for real-time response.In order to improve the quality and timeliness of community governance,we propose a work-order disposal decision-making method based on event portrait and case-based reasoning.Firstly,the address gene database of geographical names is constructed in the form of triples to obtain the spectral features in geographical names based on the unified standard address database,and the tree set is constructed to represent the hierarchical relationship between address genes with the purpose of completing and restoring the missing address elements.Secondly,in order to fully explore the local and global features of the community work-order text,the community event contained in the work-order text is extracted by the combined neural network based on BiGRU,Self-Attention,CNN and CRF models.Finally,on the basis of building the historical cases for the community events,keyword extracting is used to calculate the similarity between events based on their keywords.Comparative experimental results show that the proposed method can achieve better performance than the baseline methods.