On Event Individuation for Document-Level Information Extraction
Gantt, William, Kriz, Reno, Chen, Yunmo, Vashishtha, Siddharth, White, Aaron Steven
–arXiv.org Artificial Intelligence
As information extraction (IE) systems have grown more adept at processing whole documents, the classic task of template filling has seen renewed interest as benchmark for document-level IE. In this position paper, we call into question the suitability of template filling for this purpose. We argue that the task demands definitive answers to thorny questions of event individuation -- the problem of distinguishing distinct events -- about which even human experts disagree. Through an annotation study and error analysis, we show that this raises concerns about the usefulness of template filling metrics, the quality of datasets for the task, and the ability of models to learn it. Finally, we consider possible solutions.
arXiv.org Artificial Intelligence
Oct-20-2023
- Country:
- Asia > Singapore (0.04)
- South America
- North America
- Honduras (0.04)
- Guatemala (0.04)
- Dominican Republic (0.04)
- Central America (0.04)
- United States
- Ohio (0.04)
- New York (0.04)
- Colorado (0.04)
- Virginia > Fairfax County
- McLean (0.04)
- Oregon > Multnomah County
- Portland (0.04)
- Minnesota > Hennepin County
- Minneapolis (0.14)
- El Salvador
- San Salvador > San Salvador (0.04)
- Usulután > Usulután (0.04)
- Chalatenango > Chalatenango (0.04)
- Canada
- Quebec > Montreal (0.04)
- Newfoundland and Labrador > Labrador (0.04)
- Europe
- United Kingdom (0.04)
- Ireland (0.04)
- Croatia (0.04)
- Africa
- Uganda (0.04)
- South Africa > Western Cape
- Cape Town (0.04)
- Genre:
- Research Report (1.00)
- Industry:
- Law Enforcement & Public Safety > Crime Prevention & Enforcement (1.00)
- Health & Medicine (0.94)
- Government > Regional Government (0.93)
- Energy > Power Industry (0.67)
- Technology: