A multi-round question and answer method for customer service based on machine question and answer

By collecting user input data in real time to calculate emotion values ​​and combining this with a conversation window mechanism, the customer service response style is adjusted to mirror the user's expression. This solves the problem of rigid responses from intelligent customer service systems when emotions fluctuate, and improves user satisfaction and the effectiveness of automated services.

CN122388231APending Publication Date: 2026-07-14BEIJING HAPPY SPLENDID TECH CO LTD

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
BEIJING HAPPY SPLENDID TECH CO LTD
Filing Date
2026-05-18
Publication Date
2026-07-14

AI Technical Summary

Technical Problem

Existing intelligent customer service systems are unable to effectively soothe users' emotions when handling conversations with users experiencing emotional fluctuations, and their rigid response style leads to low user satisfaction and may escalate conflicts, thus losing the value of automated services.

Method used

By collecting user input duration, text length, number of withdrawals, and number of edits in real time, the system calculates instant input speed and sentiment value. Combined with the conversation window mechanism, it quantifies sentiment changes and adjusts the rhythm and style of customer service responses to mirror user expression, thus achieving proactive intervention.

Benefits of technology

It improved user satisfaction, reduced the probability of user emotions escalating into anger or requiring a human intervention, and enhanced the sustainability of automated services.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122388231A_ABST
    Figure CN122388231A_ABST
Patent Text Reader

Abstract

The application relates to the technical field of machine question answering, and particularly discloses a customer service multi-round question answering method based on machine question answering, which comprises the following steps: S1, detecting that a user starts a session, collecting session behaviors, and calculating an instant input speed; S2, calculating a mean value and a standard deviation of the instant input speed, setting an input speed interval, calculating an emotion value based on the number of times of withdrawal and the number of times of editing of a text content box; and S3, pre-setting a session window, calculating a change value and a mean value in the session window, and adjusting a session strategy of the current round according to the change value. According to the scheme, the input speed, the number of times of withdrawal and the number of times of editing of the user are collected, the emotion deterioration trend is dynamically perceived, and intervention is made in advance; meanwhile, based on a historical characteristic dialogue, the customer service reply is synchronized with the user expression habit in the length of a sentence segment and the symbol density, the discomfort caused by a mechanical template is reduced, and the user satisfaction and the value of automatic service are improved.
Need to check novelty before this filing date? Find Prior Art