Fraud identification method, device, electronic device and storage medium

Through the method of hierarchical information processing and knowledge graph definition of telecommunications dialogue texts, the problem of difficult telecommunications fraud in the prior art is solved, and accurate identification and effective prevention of fraud information is achieved.

CN117829290BActive Publication Date: 2025-03-25HANGZHOU CHENGZHI TIANYANG TECH CO LTD
2 Cites 0 Cited by

Patent Information

Application Number
CN202311869370.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-12-29
Publication Date
2025-03-25
Estimated Expiration
2043-12-29

AI Technical Summary

Technical Problem

It is difficult for the existing technology to effectively identify and prevent telecommunications fraud, especially when fraudsters forge numbers and use diversified fraud methods. Traditional methods cannot achieve precise prevention, resulting in frequent cases of fraud.

Method used

A fraud pre-identification method of hierarchical information is adopted to classify scenes and predict the conversation text data of voice-transformed, and combine sensitive intention prediction models to produce knowledge graphs based on caller, called and call scenes, distinguish between fraudster maps and non-fraudster maps, and identify them through the input of entity categories, scene categories and dialogue intentions.

Benefits of technology

The accurate identification of fraud information is achieved. Through the extraction of multi-dimensional text features and the definition of knowledge graphs, the error recognition rate can be effectively reduced and the accuracy and efficiency of fraud recognition can be improved.

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

Abstract

The present invention discloses a fraud identification method, device, electronic device and storage medium, which relates to information processing technology. The method comprises the following steps: inputting a scene classification and entity prediction model to output entity categories and scene categories for speech-converted dialogue text data, inputting a trained sensitive intention prediction model to output dialogue intentions for speech-converted dialogue text data; making a knowledge graph based on caller, called party and call scene data, the knowledge graph comprising a fraud sub-graph and a non-fraud sub-graph, the distinguishing features of the fraud sub-graph and the non-fraud sub-graph comprising entity category differences; inputting entity categories, scene categories and dialogue intentions into the knowledge graph, and determining that there is a suspicion of fraud if the fraud sub-graph meets the requirements. The present invention proposes a fraud pre-identification method for modeling hierarchical information, which makes use of an external knowledge base to compensate for the problem that the model cannot sample fraud samples globally.
Need to check novelty before this filing date? Find Prior Art