Correct mechanism totals and pipeline description; add card changelog

README.md

@@ -65,22 +65,28 @@ Built on the [OHCHR Universal Human Rights Index](https://uhri.ohchr.org) (UHRI)
 | Mechanism | Records | Bodies |
 |---|---|---|
 | Universal Periodic Review (UPR) | 127,625 | 1 |
-| Treaty Bodies | ~103,000 | 12 (CCPR, CAT, CERD, CEDAW, CRC, CESCR, CMW, CRPD, CED, SPT, CRC-OP-AC, CRC-OP-SC) |
-| Special Procedures (SR, IE, WG) | ~37,000 | 57 |
+| Treaty Bodies | 119,532 | 12 (CCPR, CAT, CERD, CEDAW, CRC, CESCR, CMW, CRPD, CED, SPT, CRC-OP-AC, CRC-OP-SC) |
+| Special Procedures (SR, IE, WG) | 20,514 | 57 |
+
+Totals are exact counts over this dataset's `body` field (127,625 + 119,532 + 20,514 = 267,671) and can be reproduced with `collections.Counter(ds["train"]["body"])`.
 
 **198 state parties** covered, including Cook Islands, Niue (non-UN-member states), State of Palestine\*, Holy See (UN observer states), and Kosovo\* (contested status).
 
 ## Cleaning pipeline (5 stages)
 
-| Stage | Issue | Records affected |
+The canonical pipeline description (with full detail and examples) lives on the dashboard's [Methodology page](https://lszoszk.github.io/UnitedNations_recommendations/dashboard.html#view=methodology). Summary:
+
+| Stage | What it does | Scale |
 |---|---|---|
-| 1 | OCR/encoding artefacts (ligatures, spacing, smart quotes) | ~8,000 |
-| 2 | Structural normalisation (section headings, paragraph markers) | ~3,500 |
-| 3 | Deduplication of near-identical records | ~1,200 |
-| 4 | Country backfill from UN document symbol | 106 |
-| 5 | Body name normalisation (78 → 70 unique bodies) | 1,481 |
+| 1 | Rule-based OCR & HTML repair (no AI): stray tags, mid-word OCR splits (`develop- ment` → `development`), whitespace, broken citations | Stages 1–4 together edited ~56,000 records (≈21 % of the dataset) |
+| 2 | LLM-assisted residue review for hard OCR corruption (bounded prompt, structural validation) | 412 records (0.15 %) |
+| 3 | Deterministic `annotation_type` normalisation (multilingual regex classifier, audit-tagged) | 2,603 of ~3,294 missing/UUID-coded types re-labelled |
+| 4 | Country backfill from the UN document symbol (`CEDAW/C/ALB/CO/4` → Albania) | 106 records |
+| 5 | Artefact drop — rows with no citable content (empty placeholders, duplicate annotation IDs, HTML-only scaffolding) | 11 rows (267,682 → 267,671) |
+
+`text_cleaned` reflects Stages 1–2 (text repair); Stages 3–5 act on metadata and row selection. `text_original` preserves the raw OHCHR export text.
 
-The `text_cleaned` field reflects all five stages. `text_original` preserves the raw OHCHR export text.
+**Applied data fixes (included in this parquet):** body-name normalisation merged duplicate issuing-body spellings (six treaty bodies and several Special-Procedure mandates appeared under both short codes and long names upstream) and resolved 20 blank `body` values from their document symbols — **1,481 rows changed, 78 → 70 distinct bodies**. The same fixes are folded into the live dashboard's data pipeline.
 
 ## Schema
 
@@ -144,3 +150,8 @@ Source data © OHCHR — shared freely and without restriction per the **[UHRI d
 > ⚠️ The OHCHR data policy may change. Always consult the current terms at [uhri.ohchr.org/en/our-data-api](https://uhri.ohchr.org/en/our-data-api) before redistributing. This dataset card reflects the policy as of 2026-05-18.
 
 Cleaning pipeline, normalisation, Stage 4 country backfill, and all metadata additions © L. Szoszkiewicz (Adam Mickiewicz University, Poznań), released under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) — free for research and education, commercial use requires permission.
+
+## Card changelog
+
+- **2026-06-11** — corrected the mechanism totals (Treaty Bodies 119,532; Special Procedures 20,514 — previously listed as rough estimates that were each ~16k off) and replaced the pipeline summary with the canonical five-stage description matching the published methodology.
+- **2026-05-18** — initial release.