首页|基于NLP技术的电网客服工单文本内容自动识别方法

基于NLP技术的电网客服工单文本内容自动识别方法

扫码查看
电网客服工单文本识别对于电力企业市场经营具有深远的意义,能为电网智能化推进提供必要的数据支撑.本文提出了一种基于自然语言处理的电网客服工单文本自动识别方法,利用改进的HanLP分词器分词,通过跨语言预训练模型得到文本的数据化表示,并结合注意力机制和双向长短时记忆模型对文本进行识别.实验数据表明,在工单的一级分类目录下,提出的模型测试集准确率可达98.3%,其精确率、召回率、F1分数分别为95.3%、94.7%、95.0%.在二级分类目录下其评价指标依然维持在90%以上.说明模型在电网客服工单文本识别方面性能较为优异.
Automatic Recognition of Work Order Text Content in Power Grid Customer Service Based on NLP Technology
The text recognition of power grid customer service work orders has profound significance for the market operation of power enterprises,and can provide necessary data support for the intelligent promotion of the power grid.This research proposes an automatic text recognition method for power grid customer service work orders based on natural language processing.The improved HanLP word breaker is used to segment words,and the data representation of the text is obtained through the cross language pre train-ing model,and the text is recognized by combining the attention mechanism and the two-way short-term memory model.The experi-mental data shows that under the first level classification directory of the work order,the accuracy of the proposed model test set can reach 98.3%,with accuracy,recall,and F1 scores of 95.3%,94.7%,and 95.0%,respectively.Under the secondary classification directory,its evaluation index remains above 90%.The model has excellent performance in text recognition of power grid customer service work orders.

natural language processingpower grid work ordertext recognitioncross-lingual language modelbidirectional long short-term memoryattention mechanism

张岚、王献军、张哲、董李锋、李卫卫

展开 >

国网河南省电力公司营销服务中心,郑州 450000

河南九域腾龙信息工程有限公司,郑州 450052

自然语言处理 电网工单 文本识别 跨语言预训练 双向长短时记忆 注意力机制

国家电网河南省营销服务中心短信智能卡片服务项目

B217X022J006

2024

自动化与仪器仪表
重庆工业自动化仪表研究所,重庆市自动化与仪器仪表学会

自动化与仪器仪表

CSTPCD
影响因子:0.327
ISSN:1001-9227
年,卷(期):2024.(2)
  • 13