About Tayyar

What's in the current release

• 20 MENA countries on the map, each with electoral system, government type, and political-history context.
• 98 political parties scored on 20 axes by the rater panel, each with per-field source citations.
• 274 political figures — heads of state and government, ministers, party leaders, opposition voices, activists, journalists, and historical leaders for regime-change context.
• 20 calibrated axes — economic, social, state-religion, democracy, west-alignment, regional-stance, palestinian-question, civil-liberties, regime-stance, pan-Arab, federalism, modernization, gender, iran-posture, press-freedom, sectarianism. Each carries a formal scoring rubric with anchored MENA exemplars; the full catalog is at /tayyar/issues.
• 328 semi-live events on the Pulse feed — recent and historical political shifts from the Arab Spring through 2026, with a confidence taxonomy (confirmed / reported / rumored / speculative) so in-flux developments can be tracked alongside settled facts.
• 54 MENA bills in the Bills catalog, each with full text references, status (enacted, repealed, struck down), and impact summaries.
• 426 primary-source documents in /tayyar/documents — verbatim manifestos, charters, parliamentary speeches, UN addresses — backing the document-grounded scoring layer.
• Declared vs. behavioral lens — 20 parties where the declared position diverges from the behavioral record on one or more axes, under the construct rule (declared by default; the governing record on the conduct axes for actors in power).
• Field-level citations — every party and figure carries external source citations on its facts. Verification runs an open ladder (unverified → author-verified → externally-verified), tracked on the methodology page rather than collapsed into a single "verified" stamp.

Editorial posture

• No fabricated traction. No invented endorsements. No exaggerated sample sizes.
• Methodology is the product. Every score traces back to its cited sources.
• Each score's confidence and review status is on the page, not hidden behind a footnote.
• Limitations are documented on the methodology page; treating any score as load-bearing without checking the underlying citations is the reader's call.

How the scores are made

Positions come from a multi-rater panel, not a single coder. A hand-coded author baseline seeds each cell; several language models — currently nine, from eight labs (Claude, Gemini, GPT, Gemma, Grok, Kimi, DeepSeek, Qwen, MiniMax) — then score the same cell independently and blind, against a shared, published rubric, with blind human anchors planned for a later round. The published number is the panel's median, and the spread between raters is reported as the cell's status — settled, contested, or provisional. You can read the whole panel — every rater's score and reasoning — on the ratings page.

The instrument is in its pilot phase. The multi-model rounds are landing country by country, with interim inter-rater reliability already reported and formal validation (human anchors, document grounding) still planned; what it can't yet do is written down on limitations. There is also a paste-and-position analyzer: drop in any political text and read its location on the 20 axes, with the passages that drove the placement.

Licensing and citation

The dataset is under CC BY-NC-SA 4.0; quoted source texts are held under fair use; the code will be open-sourced under MIT. For citation, see /dataset, which provides BibTeX and APA forms. Cite the frozen tayyar-dataset-v0.2 snapshot for a stable reference; a public repository and DOI are in preparation.