Multi-event extraction method, system, device and medium based on graph self-dependent network
CN115827851BActive Publication Date: 2026-02-27CHENGDU UNION BIG DATA TECH CO LTD
View PDF 0 Cites 0 Cited by
Patent Information
- Application Number
- CN202211694866.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-12-28
- Publication Date
- 2026-02-27
- Estimated Expiration
- 2042-12-28
AI Technical Summary
Technical Problem
Existing event extraction techniques cannot effectively account for dependencies between multiple events and require complex natural language processing tools.
Method used
We employ a graph autodependent network for multi-event extraction and use a deep conditional autodependent network and a graph attention network to model multi-event dependencies, directly generating global event and argument information from the detected text without the need for additional tools.
Benefits of technology
It improves the performance of multi-event extraction, has good scalability and effectiveness, can be used in different fields, and reduces the dependence on manual annotation and natural language processing tools.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN115827851B_ABST
Abstract
The application provides a multi-event extraction method, system, device and medium based on a graph self-dependent network, relates to the field of natural language processing information extraction, and the method comprises the following steps: acquiring primary prediction information; adopting a graph self-dependent network to perform multi-event dependency relationship modeling to obtain global event information; outputting trigger word text based on the primary prediction information and the global event information; adopting a conditional fusion function to integrate the trigger word text in the detection text to obtain the primary prediction information; adopting the graph self-dependent network to perform multi-argument dependency relationship modeling to obtain global argument information; and outputting event argument text based on the primary prediction information and the global argument information. The application uses the graph self-dependent network to mine the dependency between multi-events, does not need to rely on other natural language processing tools, allows learning the relevance of different trigger words and arguments, and makes the graph self-dependent network have good expansibility and effectiveness.
Need to check novelty before this filing date? Find Prior Art