Regional nationality hotspot mining and brief report generation method and system based on complex network and large language model

By constructing a temporal semantic hotspot network and a complex network community discovery algorithm, and combining it with a large language model to generate structured briefings, the dynamic and integrated problems of public opinion analysis in existing technologies are solved, and efficient, accurate analysis and full-process automation of regional and national multilingual public opinion are achieved.

CN122089296APending Publication Date: 2026-05-26SHANGHAI YIZHE INFORMATION TECH CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202610058125.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-01-16
Publication Date
2026-05-26

AI Technical Summary

Technical Problem

When faced with complex and dynamically evolving regional and national multilingual public opinion, existing technologies struggle to capture the dynamic evolution of public opinion, community discovery is disconnected from semantic themes, and information summarization capabilities are one-sided and incomplete, resulting in static and fragmented analysis results with limited accuracy and insufficient information integration.

Method used

By constructing a temporal semantic hotspot network, utilizing complex networks and large language models, a directed network is built based on temporal constraints and semantic similarity between texts. This is combined with community detection algorithms to identify hot event communities, and structured hotspot briefings are automatically generated through generative large language models.

Benefits of technology

It effectively captures the dynamic evolution of public opinion in multiple languages, improves the accuracy and completeness of hot topic discovery, enhances analysis efficiency and depth, and generates reports that naturally reflect the full picture and diverse viewpoints of events.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a regional nationality hotspot mining and brief report generation method and system based on a complex network and a large language model, and belongs to the technical field of natural language processing and public opinion analysis. The method comprises the steps that multi-language public opinion data are collected, cleaned, deduplicated and sorted; constructing a news propagation network based on the time sequence semantic similarity, and discovering a hotspot community by using a connected component algorithm; performing quality evaluation and theme consistency optimization on the community, and screening representative texts through an improved PageRank algorithm; and in combination with a preset prompt template, calling the generative large language model, and automatically generating a structured brief report comprising an event summary, a knowledge graph, news rewriting and brief comments and hotspot extension. According to the method, the dynamic evolution venation of multi-language public opinions can be effectively captured, the accuracy and integrity of hot event mining are improved, full-process automation from data acquisition to report generation is realized, and the efficiency and depth of regional nationality public opinion analysis are remarkably improved.
Need to check novelty before this filing date? Find Prior Art