A public security intelligence text element extraction method

By pre-training and knowledge augmentation of the SafeBERT model, combined with a professional vocabulary and trigger word set, a processing model suitable for public security intelligence texts is constructed. This solves the problem of insufficient efficiency and accuracy in intelligence text processing in the public security field and achieves efficient automated processing.

CN122173652APending Publication Date: 2026-06-09DACE INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
DACE INFORMATION TECH CO LTD
Filing Date
2024-11-22
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing technologies lack targeted algorithms to achieve rapid and intelligent processing of public safety intelligence text information in the public safety field, resulting in insufficient processing efficiency and accuracy.

Method used

The model is pre-trained using the SafeBERT model and combined with a public security professional vocabulary and trigger word set. Through data cleaning, knowledge augmentation and data augmentation, a model suitable for extracting elements from public security intelligence texts is constructed, including an embedding layer, SafeBERT and CRF layers. The Masked language model is used for case classification.

Benefits of technology

It improves the efficiency and security of public security intelligence text processing, replaces manual processing, and ensures the recognition rate, recall rate, and relationship accuracy of important entities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122173652A_ABST
    Figure CN122173652A_ABST
Patent Text Reader

Abstract

This invention proposes a method for extracting elements from public safety intelligence texts, comprising the following steps: collecting original public safety intelligence text data related to the public safety field, and cleaning the original public safety intelligence text data to include only specific formats of case causes and descriptions; eliminating negative examples through public safety event detection to construct a reliable domain pre-training corpus; pre-training the training corpus and adding a public safety professional vocabulary for knowledge enhancement; adding a candidate trigger word set for data enhancement; initializing the SafeBERT model, adding the pre-trained public safety corpus, and then continuing to perform MLM and case cause classification tasks to obtain a trained SafeBERT model; and constructing a model suitable for public safety intelligence text element extraction based on the SafeBERT model. This achieves computerized processing of public safety intelligence text information in the vertical field of public safety, replacing manual processing methods to improve the efficiency and security of public safety intelligence text processing.
Need to check novelty before this filing date? Find Prior Art