Add a clean en_us audio subset.

README.md

@@ -56,6 +56,37 @@ configs:
   data_files:
   - split: test
     path: "audio/*.parquet"
+- config_name: audio_en_us_clean
+  features: 
+    - name: utt_id
+      dtype: string
+    - name: waveform
+      dtype:
+        audio:
+          sampling_rate: 16000
+          decode: false
+    - name: locale
+      dtype: string
+    - name: speaker_id
+      dtype: string
+    - name: speaker_age
+      dtype: int32
+    - name: speaker_gender
+      dtype: string
+    - name: environment
+      dtype: string
+    - name: text
+      dtype: string
+    - name: topk_salient_terms
+      sequence:
+        dtype: string
+    - name: topk_salient_terms_timestamps
+      sequence:
+        sequence:
+          dtype: float32
+  data_files:
+  - split: test
+    path: "audio/utt_en_us_clean.parquet"
 - config_name: document_retrieval_cross_lang
   data_files:
   - split: test