A content topic analysis method based on a large language model

By using a large language model-based approach, features are extracted and classified from dialogue content. Combined with a network plugin to obtain background knowledge, the problem of multiple topic intersections and online hot topics is solved, and fine-grained topic segmentation and summary generation of dialogue content are achieved.

CN120144753BActive Publication Date: 2026-06-16NAT COMP NETWORK & INFORMATION SECURITY MANAGEMENT CENT
View PDF 3 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
NAT COMP NETWORK & INFORMATION SECURITY MANAGEMENT CENT
Filing Date
2024-12-31
Publication Date
2026-06-16

AI Technical Summary

⚠Technical Problem

Existing topic detection technologies are ineffective when dealing with multiple overlapping topics and fail to effectively utilize the latest online news or trending topics, making it difficult to extract fine-grained topics and provide summary overviews in a short period of time.

⚗Method used

This paper employs a large language model-based approach to preprocess dialogue content, extract various features, and combine them with the BERT model for topic classification. It then uses a sliding window strategy and bidirectional LSTM layers to cluster topics, extracts discussion objects and key phrases using the large language model, and combines network plugins to obtain background knowledge to form topic names and summaries.

🎯Benefits of technology

It achieves fine-grained topic segmentation and accurate extraction of topic summaries from dialogue content, and can handle multiple topic changes in a short time, providing personalized analysis reports.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120144753B_ABST
    Figure CN120144753B_ABST
Patent Text Reader

Abstract

The application discloses a topic analysis method based on a large language model, data preprocessing is performed on dialogue content; topic classification is performed on the preprocessed dialogue content, and content belonging to the same topic is collected together; a discussion object, an abbreviation, and a key phrase appearing in the content are extracted, and additional background knowledge is collected; a large language model is used to summarize the content under the topic, and a topic name and a topic abstract are formed; multiple topics in a dialogue process are processed in the same way, and a topic analysis report of discussion content is formed. The application can efficiently and accurately extract important information and form a topic report.
Need to check novelty before this filing date? Find Prior Art