Record Linkage for Quantifying Conflict Casualties in Syria
summary
Human Rights Data Analysis Group applies random forests and multiple systems estimation to four documentation sources — Syrian Center for statistics and research, Syrian Network for human rights, the Syrian shua website, and the Violation Documentation Center — performing record linkage on name date location matching with a native Arabic reviewer to label training pairs, then clustering and merging to produce a single unique-victim dataset. Models then estimate the large unseen count of unreported victims and map documentation coverage across Syria, showing wide geographic variability.
AI-written summary · about
chapters (7)
- 0:00Start
- 3:51Conflicting Narratives
- 4:27Event Size Bias
- 5:08Documented Identifiable Victims
- 6:26The Syrian Center for Statistics and Research
- 8:01Documentation Patterns
- 8:38Documentation Landscape
related talks
01mMachine Learning for Human Rights Advocacy
06mLeveraging Free-Hand Sketches for Potential Screening of PTSD
24mKeynote: Digital Echoes: Understanding Mass Violence with Data and Statistics
01h 43mPerspective
28mData Science for Social Good: Making Impact in Resource-Constrained Environments
17mCombating PTSD with Machine Learning
30mData Driven Security
31mValidating Vulnerability Analysis with Statistical Analysis of Metadata