Uncovering policy priorities for disability inclusion: NLP and LLM approaches to analyzing CRPD state reports
Bibliographic record
Abstract
Abstract Over 193 countries have signed at least one of more than 500 multilateral treaties addressing critical global issues, such as human rights, environmental protection, and trade. Ratifying a treaty obligates a country, as a “State Party,” to report to the United Nations on its progress toward implementing the treaty’s provisions. These reports and their associated review processes generate a wealth of textual data. Effectively monitoring, reviewing, and assessing national, regional, and global progress toward these treaty commitments is crucial for ensuring compliance and realizing the benefits of international cooperation. The UN Convention on the Rights of Persons with Disabilities (CRPD), which has been ratified by 191 countries, exemplifies this challenge. With over 1.3 billion people worldwide living with disabilities, the CRPD aims to promote a shift from a charity-based “medical model” that views disability as an individual deficiency, to a rights-based “social justice model” that emphasizes societal barriers and inclusivity. Each State Party submits periodic reports to the Committee on the Rights of Persons with Disabilities detailing their implementation efforts. This study analyzed all available CRPD State Reports ( N = 170) using text mining, Natural Language Processing, and GenerativeAI tools to assess global progress, identify regional variations, and explore the factors influencing successful implementation. The findings reveal evidence of widespread CRPD implementation, growing support for social justice and economic inclusion, and the importance of civil society engagement. Hybrid data analysis approach of this study offers a promising framework for harnessing the power of textual data to advance the realization of treaty commitments worldwide.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.022 | 0.092 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.026 | 0.024 |
| Science and technology studies | 0.002 | 0.002 |
| Scholarly communication | 0.008 | 0.004 |
| Open science | 0.002 | 0.005 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".