Synchronize YAML configs and custom splits; preserve prose

README.md

@@ -0,0 +1,54 @@
+---
+configs:
+- config_name: unicode
+  data_files:
+  - split: train_19d
+    path: subsets/unicode/train_19d-*.parquet
+  - split: train_22d
+    path: subsets/unicode/train_22d-*.parquet
+  - split: train_24d
+    path: subsets/unicode/train_24d-*.parquet
+  - split: train_25d
+    path: subsets/unicode/train_25d-*.parquet
+  - split: train_29d
+    path: subsets/unicode/train_29d-*.parquet
+  - split: train_500d
+    path: subsets/unicode/train_500d-*.parquet
+  - split: train_2500d
+    path: subsets/unicode/train_2500d-*.parquet
+- config_name: wordnet_eng
+  data_files:
+  - split: train_19d
+    path: subsets/wordnet_eng/train_19d-*.parquet
+  - split: train_22d
+    path: subsets/wordnet_eng/train_22d-*.parquet
+  - split: train_24d
+    path: subsets/wordnet_eng/train_24d-*.parquet
+  - split: train_25d
+    path: subsets/wordnet_eng/train_25d-*.parquet
+  - split: train_29d
+    path: subsets/wordnet_eng/train_29d-*.parquet
+  - split: train_500d
+    path: subsets/wordnet_eng/train_500d-*.parquet
+  - split: train_2500d
+    path: subsets/wordnet_eng/train_2500d-*.parquet
+---
+# Geometric Vocabulary — Unified
+
+This dataset aggregates multiple subsets such as Unicode and WordNet languages. Each subset is a dataset configuration. Each dimension is exposed as a split.
+
+Last updated: 2025-09-02 06:01:53 UTC
+
+Subsets and available split names
+
+| Subset | Splits | Example pattern |
+|-------:|:-------|:----------------|
+| unicode | train_19d, train_22d, train_24d, train_25d, train_29d, train_500d, train_2500d | subsets/unicode/train_19d-*.parquet |
+| wordnet_eng | train_19d, train_22d, train_24d, train_25d, train_29d, train_500d, train_2500d | subsets/wordnet_eng/train_19d-*.parquet |
+
+Loading example
+
+from datasets import load_dataset
+repo = 'AbstractPhil/geometric-vocab-all'
+ds = load_dataset(repo, 'unicode', split='train_1024d')
+print(ds)