Datasets:
docs(card): place Trackio snapshot after introduction
Browse files
README.md
CHANGED
|
@@ -67,6 +67,26 @@ Maintained by **Noé Flandre**.
|
|
| 67 |
|
| 68 |
Source code: [GitHub repository](https://github.com/NoeFlandre/osm-polygon-wikidata-only).
|
| 69 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
## Coverage
|
| 71 |
|
| 72 |
### Polygons with Wikipedia or Wikivoyage text
|
|
@@ -334,26 +354,6 @@ This table is recomputed from the finalized Parquet tables before each dataset-c
|
|
| 334 |
- `Polygons with Wikipedia or Wikivoyage text`: polygons satisfying the Wikipedia condition or sharing a Wikidata entity with at least one non-empty Wikivoyage document. Each polygon is counted once.
|
| 335 |
- `Text coverage`: `combined text-covered polygons / all dataset polygons` in the continent row.
|
| 336 |
|
| 337 |
-
## Trackio snapshot
|
| 338 |
-
|
| 339 |
-
The finished dataset is recorded in the single public Trackio run [`final-dataset-snapshot`](https://huggingface.co/spaces/NoeFlandre/osm-polygon-wikidata-only-trackio). It contains exactly three plots: a text-coverage funnel, the top ten Wikipedia languages plus `Other languages`, and dataset composition on a logarithmic scale. The run is a static snapshot with no pipeline timeline.
|
| 340 |
-
|
| 341 |
-
| Snapshot metric | Value |
|
| 342 |
-
| --- | ---: |
|
| 343 |
-
| Polygons | 1,184,110 |
|
| 344 |
-
| Wikipedia + Wikivoyage documents | 2,288,170 |
|
| 345 |
-
| Document words | 801,528,334 |
|
| 346 |
-
| Languages | 351 |
|
| 347 |
-
| Geographic regions | 375 |
|
| 348 |
-
|
| 349 |
-
| Small snapshot table | Value |
|
| 350 |
-
| --- | ---: |
|
| 351 |
-
| Wikipedia polygon-document links | 2,468,604 |
|
| 352 |
-
| Polygon/link-table storage | 9.9 GB |
|
| 353 |
-
| Total Parquet storage | 19.2 GB |
|
| 354 |
-
|
| 355 |
-
The funnel's language thresholds use the canonical Wikipedia polygon fields. Wikivoyage is included in the combined document, word, and dataset-composition totals.
|
| 356 |
-
|
| 357 |
## Schema
|
| 358 |
|
| 359 |
### `polygons`
|
|
|
|
| 67 |
|
| 68 |
Source code: [GitHub repository](https://github.com/NoeFlandre/osm-polygon-wikidata-only).
|
| 69 |
|
| 70 |
+
## Trackio snapshot
|
| 71 |
+
|
| 72 |
+
The finished dataset is recorded in the single public Trackio run [`final-dataset-snapshot`](https://huggingface.co/spaces/NoeFlandre/osm-polygon-wikidata-only-trackio). It contains exactly three plots: a text-coverage funnel, the top ten Wikipedia languages plus `Other languages`, and dataset composition on a logarithmic scale. The run is a static snapshot with no pipeline timeline.
|
| 73 |
+
|
| 74 |
+
| Snapshot metric | Value |
|
| 75 |
+
| --- | ---: |
|
| 76 |
+
| Polygons | 1,184,110 |
|
| 77 |
+
| Wikipedia + Wikivoyage documents | 2,288,170 |
|
| 78 |
+
| Document words | 801,528,334 |
|
| 79 |
+
| Languages | 351 |
|
| 80 |
+
| Geographic regions | 375 |
|
| 81 |
+
|
| 82 |
+
| Small snapshot table | Value |
|
| 83 |
+
| --- | ---: |
|
| 84 |
+
| Wikipedia polygon-document links | 2,468,604 |
|
| 85 |
+
| Polygon/link-table storage | 9.9 GB |
|
| 86 |
+
| Total Parquet storage | 19.2 GB |
|
| 87 |
+
|
| 88 |
+
The funnel's language thresholds use the canonical Wikipedia polygon fields. Wikivoyage is included in the combined document, word, and dataset-composition totals.
|
| 89 |
+
|
| 90 |
## Coverage
|
| 91 |
|
| 92 |
### Polygons with Wikipedia or Wikivoyage text
|
|
|
|
| 354 |
- `Polygons with Wikipedia or Wikivoyage text`: polygons satisfying the Wikipedia condition or sharing a Wikidata entity with at least one non-empty Wikivoyage document. Each polygon is counted once.
|
| 355 |
- `Text coverage`: `combined text-covered polygons / all dataset polygons` in the continent row.
|
| 356 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 357 |
## Schema
|
| 358 |
|
| 359 |
### `polygons`
|