Upload docs/data_dictionary.md with huggingface_hub

docs/data_dictionary.md

@@ -10,11 +10,21 @@ Column-level schemas for every CSV in `data/` and `samples/`.
 | `name` | text | Speaker name (public figure). | celebrity name |
 | `group` | int | Sociophonetic group code. | 1-5 (see below) |
 | `gender` | int | Speaker gender. | 1=male, 2=female |
-| `age` | int | Speaker age bracket. | 1=younger, 2=older |
+| `age` | int | Speaker age bracket. | 1=under 45, 2=55 or older |
 
 ### Sociophonetic group mapping
 
-The 5 sociophonetic groups partition the 100 speakers into 20 speakers each. The integer-to-group mapping is documented by the dataset authors. The benchmark protocol uses `group` only as a stratification variable for fairness analyses; downstream users reproducing fairness analyses can rely on the integer codes directly.
+The 5 sociophonetic groups partition the 100 speakers into 20 speakers each:
+
+| Code | Group |
+|---|---|
+| 1 | New York City English |
+| 2 | Southern American English |
+| 3 | African American English |
+| 4 | Latino English |
+| 5 | Asian American English |
+
+The benchmark protocol uses `group` as a stratification variable for cross-validation folds (gender-balanced, speaker-level) and for fairness analyses.
 
 ## `stimuli.csv` (9,800 rows)