一种客服话术生成方法、设备及介质
By acquiring questions and historical conversations from the customer service system, constructing context samples using backtracking steps, and conducting quality assessments, high-quality customer service scripts are generated. This solves the problem of inflexible interaction in existing technologies and enables more accurate and context-aware responses.
CN116049364BActive Publication Date: 2026-07-17SHANDONG SYNTHESIS ELECTRONICS TECH
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG SYNTHESIS ELECTRONICS TECH
- Filing Date
- 2022-12-30
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing methods for obtaining customer service scripts suffer from inflexible interaction, making it difficult to generate high-quality and context-appropriate responses.
Method used
By acquiring questions and historical dialogues, context sample pairs are extracted using a pre-set backtracking step size, an input set is constructed, and the input is fed into the speech generation network. A loss metric network is used for quality evaluation to determine the optimal answer.
Benefits of technology
It improved the accuracy and contextual relevance of customer service scripts, avoided issues of non-standard and incoherent language, and enhanced the quality of interaction.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116049364B_ABST
Abstract
本申请公开了一种客服话术生成方法、设备及介质，方法包括：获取问题和历史对话，根据预先设置的回溯步长在历史对话抽取上下文样本对，并根据上下文样本对和问题构建输入集；将输入集输入话术生成网络，以得到客服话术；将客服话术输入损失度量网络，以得到特征对；将特征对带入预先设置的质量函数，以得到质量因子，并将质量因子与预先设置的相似度阈值进行比较，若质量因子大于相似度阈值，则将客服话术标记为最优答案。本申请有效提升了准确率，同时避免了话术生成方法存在的语言不规范、不通顺等问题。
Need to check novelty before this filing date? Find Prior Art