A risk public opinion analysis method based on natural language processing

CN120994833BActive Publication Date: 2026-06-19CHANGZHOU JIADO HIGH-TECH CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
CHANGZHOU JIADO HIGH-TECH CO LTD
Filing Date
2025-07-24
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing natural language processing technologies struggle to effectively handle heterogeneous data and identify the propagation paths and causal relationships of risk events in public opinion analysis. Furthermore, they lack cross-language capabilities and the ability to protect sensitive information, resulting in inaccurate risk warnings.

Method used

Hash pseudo-identifiers are used for de-identification, a dual-classification weak supervision framework is constructed, risk confidence is generated and mapped to a dynamic heterogeneous risk knowledge graph, risk propagation analysis is performed through a multi-head spatiotemporal diffusion Transformer model, and risk warnings are pushed to the back-end control center through a causal comparison analysis model to output an early risk warning index.

Benefits of technology

It enables secure collection and efficient risk identification of heterogeneous multi-source data, cross-language semantic fusion, and dynamic risk propagation analysis, thereby improving the automation, accuracy, and reliability of public opinion risk monitoring. It has high security and universality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120994833B_ABST
    Figure CN120994833B_ABST
Patent Text Reader

Abstract

This invention discloses a risk-based public opinion analysis method based on natural language processing, comprising: real-time capture of public opinion text; desensitizing user identity fields using hash pseudo-identifiers to obtain the original text stream; constructing a dual-classification weakly supervised framework to calculate risk confidence; when the risk confidence is ≥ a first threshold, writing the corresponding text into a labeling pool to form a risk corpus; mapping the risk corpus into a dynamic heterogeneous risk knowledge graph according to entity-relationship-time slice meta-path rules; applying a multi-head spatiotemporal diffusion Transformer to the subgraphs intersecting with the current window in the graph to obtain the risk propagation vector of the event node; inputting the risk propagation vector and the historical baseline vector into a causal comparison analysis model to output an early risk warning index; when the warning index is ≥ a second threshold, pushing a risk alarm containing the event path to the backend control center. This invention effectively improves the automation, accuracy, and reliability of public opinion risk monitoring and early warning.
Need to check novelty before this filing date? Find Prior Art