Update README.md

README.md

@@ -20,13 +20,13 @@ dataset_info:
           '6': anger
   splits:
   - name: train
-    num_bytes: 26772110.0
+    num_bytes: 26772110
     num_examples: 304
   - name: test
-    num_bytes: 20866781.0
+    num_bytes: 20866781
     num_examples: 231
   download_size: 46818101
-  dataset_size: 47638891.0
+  dataset_size: 47638891
 configs:
 - config_name: default
   data_files:
@@ -34,4 +34,30 @@ configs:
     path: data/train-*
   - split: test
     path: data/test-*
+task_categories:
+- audio-classification
+tags:
+- audio
+- paralinguistic
+- multiclass
 ---
+
+# EmoDB
+
+The EmoDB is the freely available German emotional database, containing a total of 535 utterances.
+It comprises of seven emotions: 1) anger; 2) boredom; 3) anxiety; 4) happiness; 5) sadness; 6) disgust; and 7) neutral.
+The data was recorded at a 48-kHz sampling rate and then down-sampled to 16-kHz.
+We follow the unofficial speaker-independent train/test split from [here](https://github.com/audeering/emodb/blob/master/CHANGELOG.md).
+
+## Citations
+
+```bibtex
+@inproceedings{burkhardt2005database,
+  title={A database of German emotional speech.},
+  author={Burkhardt, Felix and Paeschke, Astrid and Rolfes, Miriam and Sendlmeier, Walter F and Weiss, Benjamin and others},
+  booktitle={Interspeech},
+  volume={5},
+  pages={1517--1520},
+  year={2005}
+}
+```