基于BM25、文本Embeddings与交叉编码器的民航客服知识库检索研究  

Research on civil aviation customer service knowledge base retrieval based on BM25,text vector method and cross encoder

作　　者：郑少帅 翁境鸿 蒋小洋 Zheng Shaoshuai;Weng Jinghong;Jiang Xiaoyang(Xiamen Airlines Co.,Ltd.,Xiamen 361006,China)

机构地区：[1]厦门航空有限公司,福建厦门361006

出　　处：《无线互联科技》2023年第24期122-125,共4页Wireless Internet Technology

摘　　要：随着民航经济的发展和人民生活水平的提高,旅客出行的服务要求越来越高,而当前传统的民航客服知识库检索普遍存在检索准确率以及效率低的问题,已经不能满足旅客的服务需求。文章通过结合Best Match 25算法、文本Embeddings和交叉编码器对知识库进行检索,高效检索出符合座席意图的答案,进而提升民航客服知识库查找效率,缩短座席通话查询时长,提升旅客服务体验,助力实现民航客服数字化、智能化转型。With the development of civil aviation economy and the improvement of people's living standards,the service requirements for passenger travel are becoming higher and higher.However,the traditional civil aviation customer service knowledge base retrieval generally has low retrieval accuracy and efficiency,which can no longer meet the service needs of passengers.This article combines the Best Match 25 algorithm,text Embeddings,and Cross-Encoders to search the knowledge base,efficiently searching for answers that match the intentions of the seat,thereby improving the efficiency of civil aviation customer service knowledge base search,shortening the call query time of the seat,improving the passenger service experience,and laying a solid foundation for the digital and intelligent transformation of civil aviation customer services.

关 键 词：民航客服 Best Match 25算法 文本Embeddings 交叉编码器 座席意图 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]