Memory enhancement and reasoning method and system for intelligent interaction

CN121998106BActive Publication Date: 2026-06-19SHANGHAI MINGQI NETWORK TECH CO LTD

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Patents(China)
Current Assignee / Owner: SHANGHAI MINGQI NETWORK TECH CO LTD
Filing Date: 2026-04-10
Publication Date: 2026-06-19

AI Technical Summary

Technical Problem

Existing intelligent interaction methods lack effective use of historical interaction information when processing user interaction data. This makes it difficult to accurately understand the user's true intentions when faced with ambiguous or unclear user expressions or complex interaction scenarios, thus affecting the accuracy and fluency of the interaction.

Method used

By capturing the original user interaction data stream generated by the intelligent interactive terminal, the interaction events are segmented and arranged in sequence, the semantics of the interaction content and the operation type are analyzed, interaction intent tags are generated, and contextual knowledge encoding and semantic fusion are performed using a memory-enhanced reasoning network. Historical interaction memory entries are retrieved to complete the current interaction intent.

Benefits of technology

Effective integration of interactive information improves the accuracy and fluency of intent analysis, enables a deep understanding of user intent and accurate responses, and enhances the accuracy of intelligent interaction and user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN121998106B_ABST

Patent Text Reader

Abstract

This application provides a memory enhancement and reasoning method and system for intelligent interaction, belonging to the field of intelligent interaction technology. First, it captures the user's original interaction data stream, segments and arranges it into a temporal sequence of interaction events, initially determines the interaction intent, and merges it into an interaction round unit sequence. Next, it extracts core keywords, associates and expands them to obtain background supporting knowledge units, generating an interaction contextual knowledge set. Then, it fuses the contextual knowledge encoding vector with the semantics of the interaction content, retrieving historical memory entries from a long-term interaction memory bank. Finally, it completes the current interaction intent based on historical memories, generates response reasoning content, and delivers it to the intelligent interaction terminal for presentation. This invention improves the accuracy of intelligent interaction and the user experience.

Need to check novelty before this filing date? Find Prior Art