Update README.md

README.md

@@ -36,10 +36,10 @@ dataset_info:
     num_bytes: 798310802.768
     num_examples: 4548
   - name: test
-    num_bytes: 70180067.0
+    num_bytes: 70180067
     num_examples: 397
   - name: validation
-    num_bytes: 61840955.0
+    num_bytes: 61840955
     num_examples: 335
   download_size: 839882578
   dataset_size: 930331824.768
@@ -52,4 +52,17 @@ configs:
     path: data/test-*
   - split: validation
     path: data/validation-*
+language:
+- sw
 ---
+
+
+Small dataset derived from Common Voice 20.0 (cv-corpus-22.0-2025-06-20) by selecting only utterances with minimum upvote counts.
+This dataset is derived from Mozilla's Common Voice 20.0 (cv-corpus-22.0-2025-06-20), see also https://commonvoice.mozilla.org. Common Voice  released under Creative Commons Zero licence (CC0 1.0).
+**Users must not attempt to determine the identity of speakers in this dataset, as per Common Voice usage terms.**
+
+This is a subset, derived by filtering based on up/down-votes. Here are the resulting split sizes as well as filter settings:
+
+* training set: 4548 utterances, 149 speakers (0 down votes, > 2 upvotes)
+* test set: 397 utterances, 220 speakers (0 down votes, > 5 upvotes)
+* validation set:  335 utterances, 78 speakers (0 down votes, > 5 upvotes)