基于语义识别的服务热线工单智能分类研究
毛晓蕾
Research on Intelligent Classification of Service Hotline Work Orders Based on Semantic Recognition
MAO Xiaolei
-
作者信息:上海申通地铁集团有限公司运营管理部， 201103， 上海
-
Affiliation:Operation Management Department of Shanghai Shentong Metro Group Co., Ltd., 201103, Shanghai, China
-
关键词:
-
Key words:
-
DOI:10.16037/j.1007-869x.2025.05.033
-
中图分类号/CLCN:U293.2
-
栏目/Col:运营管理
摘要:
［目的］传统轨道交通网络服务热线以人工客服接听、手工填写工单和处理分类为主，客运服务人员承担着高强度、超负荷的服务工作，同时服务品质也难以得到保障。因此，有必要引入基于深度学习的语义识别技术，以实现运营管理的数字化和智能化。［方法］分析了当前服务热线的业务分类及工单智能分类系统需求，采用语义分析的分词逻辑，通过建立关键字库来提升识别的精度；将关键词库作为领域词典，构建了一种以分布式文本向量表示为基础，融合Transformer自注意力机制的智能化文本分类模型。在所提智能化文本分类模型中，注意力焦点更集中于与分类任务相关性强的词汇，降低了上下文信息中无关词汇对分类结果的干扰，并能够动态地显示在不同语境中含有不同语义的文本，实现对乘客意图的分类。在此基础上，搭建了热线工单（以下简称“工单”）记录文件智能分类系统。［结果及结论］由真实数据集上的试验结果可知，所提智能化文本分类模型具有一定的有效性与正确性。采用高模块化的软件系统设计，实现了工单的自动分类，能够有效提升工单响应速度，降低人力、物力的成本投入。所提智能化文本分类模型能够提高运营整体的服务质量和乘客满意度。
Abstracts:
［Objective］Traditional rail transit network service hotline mainly relies on manual customer service answering calls, manually filling out work orders and handling classifications. Passenger service staff undertake high-intensity and overloaded service work, while the service quality is difficult to be guaranteed. Therefore, it is necessary to introduce semantic recognition technology based on deep learning to achieve digitalized and intelligent operation management. ［Method］The systematic requirements for the business classification of current service hotline and the intelligent classification of the work order are analyzed. The word- segmentation logic of semantic analysis is used and the recognition accuracy is improved by establishing a keyword library. Using such library as a domain dictionary, an intelligent text classification model based on distributed text vector representation and integrating the Transformer self-attention mechanism is constructed. In the proposed intelligent text classification model, the attention focus is more concentrated on the words strongly relevant to the classification task, thus reducing the interference of irrelevant words in the context on the classification results, and texts with different semantics in different contexts can also be dynamically displayed to achieve the classification of passengers′ intentions. On this basis, an intelligent classification system for hotline work order records is built. ［Result & Conclusion］The experimental results on the real datasets show that the proposed intelligent text classification model has certain effectiveness and correctness. By adopting a highly modular software system design, the automatic classification of work orders is realized, which can effectively improve the work order response speed and reduce the costs of human and material resources. The proposed intelligent text classification model can improve the overall operation service quality and passenger satisfaction.
- 上一篇: 全自动运行线路多岗位联动平台的建模与仿真
- 下一篇: 运营条件下U型梁高架地铁线路环境振动实测分析
过刊浏览
资料下载
友情链接