电力信息通信客服机器人特定语义数据检索优化  

Optimization of specific semantic data retrieval for power information communication customer service robot

作　　者：王丽 蒋明 王伟 涂冰花 周明 WANG Li;JIANG Ming;WANG Wei;TU Binghua;ZHOU Ming(Information and Communication Branch of State Grid Anhui Electric Power Company,Hefei 230061,China)

机构地区：[1]国网安徽省电力有限公司信息通信分公司,安徽合肥230061

出　　处：《电子设计工程》2024年第20期168-171,176,共5页Electronic Design Engineering

基　　金：国家电网有限公司科技项目资助(SGGSPX00HLWJS2200097)。

摘　　要：针对电力信息通信客服机器人在服务客户时,受语音信号频谱偏差造成语义理解错误,导致反馈的答案内容与客户预期不一致的问题,研究电力信息通信客服机器人特定语义数据检索优化方法。通过采集客户语音信号并提取梅尔倒谱系数,以此为输入,利用神经网络实现语义识别,得出语义关键词。将提取的关键词作为答案文本描述,计算语义关键词与每个答案文本之间的匹配度,以此实现特定语义数据检索优化。结果表明,所研究检索方法的约登指数为9.625,检索时间为24.32 s,由此说明该检索方法对特定语义的准确性更高,检索效率更快。Aiming at the problem of semantic understanding errors caused by speech signal spectrum deviation when serving customers by power information communication customer service robots,which results in inconsistent feedback answers with customer expectations,this paper studies a specific semantic data retrieval optimization method for power information communication customer service robots.By collecting customer voice signals were collected and Mel Cepstrum Coefficients were extracted as input.Neural networks were used to achieve semantic recognition and obtain semantic keywords.Using the extracted keywords as answer text descriptions,calculate the matching degree between semantic keywords and each answer text to achieve specific semantic data retrieval optimization.The results show that the Jordan index of the research retrieval method is 9.625,and the retrieval time is 24.32 s,which indicates that the retrieval method has higher accuracy for specific semantics and faster retrieval efficiency.

关 键 词：电力信息通信 客服机器人 特定语义识别 匹配度 数据检索优化 

分 类 号：TN141[电子电信—物理电子学]