Related Experiment Video
Updated: Jul 15, 2026

Generating Strictly Controlled Stimuli for Figure Recognition Experiments
Published on: March 18, 2019
TGIN: Document-level event extraction with two-phase graph inference network.
This study introduces a Two-phase Graph Inference Network (TGIN) for document-level event extraction. TGIN enhances precision by explicitly modeling key information and reducing noise from irrelevant entities.
Area of Science:
- Natural Language Processing
- Information Extraction
- Artificial Intelligence
Background:
- Document-level event extraction is challenging due to scattered entities across multiple sentences.
- Existing methods struggle with implicit information modeling and irrelevant entity consideration, leading to noise and reduced efficiency.
- Effective modeling of entity interactions is crucial for accurate document-level event extraction.
Purpose of the Study:
- To propose a novel Two-phase Graph Inference Network (TGIN) for improved document-level event extraction.
- To address limitations of previous methods by explicitly modeling key information and reducing irrelevant entity influence.
- To enhance the efficiency and precision of extracting event records from entire documents.
Main Methods:
- Constructing a heterogeneous document-level graph in the first phase to capture complex interactions and acquire document-aware features.
- Employing a key information aggregator with an attention mechanism to explicitly aggregate key sentences for entity pairs.
- Building an entity-level graph in the second phase using predicted entity links as prior information to model interactions between related entity pairs.
Main Results:
- The proposed TGIN framework demonstrates superior performance on the ChFinAnn dataset for document-level event extraction.
- Explicitly aggregating key information and focusing on relevant entity pairs significantly improved extraction precision.
- The two-phase graph inference approach effectively reduced error propagation and noise in event extraction.
Conclusions:
- The TGIN approach offers a significant advancement in document-level event extraction by effectively addressing noise and efficiency issues.
- Explicitly modeling key information and entity interactions through a two-phase graph inference network leads to superior extraction performance.
- This framework provides a robust solution for extracting complex event records from unstructured documents.
More Related Videos
10:44Inherent Dynamics Visualizer, an Interactive Application for Evaluating and Visualizing Outputs from a Gene Regulatory Network Inference Pipeline
Published on: December 7, 2021
07:35A Knowledge Graph Approach to Elucidate the Role of Organellar Pathways in Disease via Biomedical Reports
Published on: October 13, 2023
Related Concept Videos
Phase Transitions
Phase Transitions: Sublimation and Deposition
Phase Diagrams
Ogive Graph
Phase Diagram
Block Diagram Reduction
The first step in this process is the identification and relocation of a branch point. A branch point, where a...