Add task category and link to paper

This PR adds the `audio-classification` task category and a link to the paper [SemantiCodec: An Ultra Low Bitrate Semantic Audio Codec for General Sound](https://huggingface.co/papers/2405.00233).

README.md

@@ -25,9 +25,11 @@ configs:
     path: data/train-*
   - split: test
     path: data/test-*
+task_categories:
+- audio-classification
 ---
 
-This dataset contains tokenized audio from [lewtun/music_genres](https://huggingface.co/datasets/lewtun/music_genres) using [SemantiCodec](https://arxiv.org/abs/2405.00233) for performing experiments on AR music generation
+This dataset contains tokenized audio from [lewtun/music_genres](https://huggingface.co/datasets/lewtun/music_genres) using [SemantiCodec](https://huggingface.co/papers/2405.00233) for performing experiments on AR music generation
 
 The following script is used for tokenization
 ```python