Publish test split with Audio() feature

README.md

@@ -1,8 +1,10 @@
 ---
 pretty_name: KMR-Tatoeba (Northern Kurdish Test Set)
-dataset_summary: >
-  Test-only speech dataset for Northern Kurdish (Kurmanji), sourced from Tatoeba.
-  Each sample includes audio, anonymized speaker ID, Kurdish transcription, and English translation.
+dataset_summary: 'Test-only speech dataset for Northern Kurdish (Kurmanji), sourced
+  from Tatoeba. Each sample includes audio, anonymized speaker ID, Kurdish transcription,
+  and English translation.
+
+  '
 language:
 - kmr
 license: cc-by-4.0
@@ -11,6 +13,27 @@ size_categories:
 task_categories:
 - automatic-speech-recognition
 - audio-text-to-text
+dataset_info:
+  features:
+  - name: audio
+    dtype: audio
+  - name: speaker
+    dtype: string
+  - name: kurdish
+    dtype: string
+  - name: english
+    dtype: string
+  splits:
+  - name: test
+    num_bytes: 150758589.0
+    num_examples: 537
+  download_size: 126408297
+  dataset_size: 150758589.0
+configs:
+- config_name: default
+  data_files:
+  - split: test
+    path: data/test-*
 ---
 
 # kmr-tatoeba

data/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d07aab7a6a0a3a976a2576ece838f6d54f4ae394bf96010153d0ea1431e94e7
+size 126408297