一种基于动态衰减的对话上下文管理方法及系统

By introducing a dual decay mechanism of time and frequency, and dynamically generating a moderating factor based on the current emotional intensity, the context management capability of the dialogue system is fundamentally improved.

CN121706805BActive Publication Date: 2026-07-17GUANGZHOU TAIDONG TECH CO LTD

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
GUANGZHOU TAIDONG TECH CO LTD
Filing Date
2025-12-17
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing dialogue systems suffer from storage bloat in long-term, multi-turn dialogue scenarios, are unable to simulate human forgetting mechanisms, lack dynamic decay over time and consideration of emotional intensity, resulting in inaccurate memory management and impacting user experience.

Method used

A dialogue context management method based on dynamic decay is adopted. Through a dual decay mechanism of time and frequency, a moderating factor is dynamically generated in combination with the current emotion intensity, the activation weight of memory items is calculated, a dialogue context vector is generated, and the ability of the dialogue management system is optimized.

Benefits of technology

This has improved the intelligence level of the dialogue system, significantly optimized the storage expansion and processing latency of the dialogue management system, and enhanced user satisfaction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121706805B_ABST
    Figure CN121706805B_ABST
Patent Text Reader

Abstract

本发明涉及人工智能交互领域的一种基于动态衰减的对话上下文管理方法和系统。首先获取包含当前情绪强度的对话状态及历史记忆项集合;接着对每个记忆项分别计算时间衰减分量和频率衰减分量;进而基于当前情绪强度生成动态调节因子,利用该因子动态调整对时间维度的时效性与频率维度的习惯性的侧重比例,经加权融合得到最终激活权重;最后基于激活权重生成对话上下文向量。本发明通过引入双重衰减与情绪调节机制,实现了类人的选择性遗忘与情境感知能力,显著提升了复杂语境下的交互智能度。
Need to check novelty before this filing date? Find Prior Art