摘要
本文基于自然语言处理、语义相似度和实体识别等算法,构建了面向网络投诉工单的智能语义自动稽核系统,挖掘回单中自然语言的命名实体、事件和关系,通过特征提取、模型构建、语义泛化和匹配度计算等环节,输出质检所需的语义关键信息,实现对网络投诉工单传统质检方法的智能化提升。同时,本文设计和实施了一种投诉工单专家规则泛化词构建方法,实现对有限专家规则特征词的补充,提升语义质检的泛化性能。通过与传统关键字匹配算法进行比较,本方法在识别精准度方面效果更优,有效降低运营商投诉工单重派率,提升了客户满意度。
Base on the algorithm of natural language processing, semantic similarity and entity identification, an intelligent semantic automatic inspecting system for network complaint work-order is constructed in this paper. It can mine named entities, incidents and relationships of natural language in receipts. Trough feature extraction, model construction, semantic generalization and matching degree calculation, the semantic key information needed for quality inspection is output, and intelligence improvement of traditional quality inspection method for network complaint work order is realized. Meanwhile, this paper designs and implements a method to construct the generalization of expert rules for complaint work orders, it can complement the feature words grown out of limited expert rules and improve the generalization performance of semantic quality inspection. Compared with the traditional keyword matching algorithm, this method has better recognition accuracy, it can effectively reduce the re-dispatch rate of complaints from operators and improve customer satisfaction.
作者
黄堃
赵东明
HUANG Kun;ZHAO Dong-ming(China Mobile Group Tianjin Co.,Ltd.,Tianjin 300020,China)
出处
《电信工程技术与标准化》
2021年第7期45-49,共5页
Telecom Engineering Technics and Standardization
关键词
自然语言处理
神经网络
语义相似度
模型泛化
实体识别
natural language processing
neural network
semantics similarity
model generalization
entity identifi cation