Update README

README.md

@@ -18,6 +18,10 @@ task_ids:
   - [Dataset Description](#dataset-description)
     - [Dataset Summary](#dataset-summary)
     - [Supported Tasks](#supported-tasks-and-leaderboards)
+  - [Dataset Structure](#dataset-structure)
+    - [Data Instances](#data-instances)
+    - [Data Fields](#data-fields)
+    - [Data Splits](#data-splits)
 
 ## Dataset Description
 
@@ -34,4 +38,30 @@ The dataset mainly contains guitar music from western classical composers, such
 
 ### Supported Tasks and Leaderboards
 
-Anyone interested can use the dataset to train a model for symbolic music generation, which consists in treating symbols for music sounds (notes) as text tokens. Then, one can implement a generative model using NLP techniques, such as Transformers.
+Anyone interested can use the dataset to train a model for symbolic music generation, which consists in treating symbols for music sounds (notes) as text tokens. Then, one can implement a generative model using NLP techniques, such as Transformers.
+
+## Dataset Structure
+
+### Data Instances
+
+Each guitar piece is represented as a string of text that contains a series of tokens, for instance:
+
+PIECE_START: Where the piece begins
+PIECE_ENDS: Where the piece ends
+TIME_SIGNATURE: Time signature for the piece
+BPM: Tempo of the piece
+BAR_START: Begining of a new bar
+NOTE_ON: Start of a new musical note specifying its MIDI note number
+TIME_DELTA: Duration until the next event
+NOTE_OFF: End of musical note specifying its MIDI note number
+
+```
+{
+  'text': PIECE_START TIME_SIGNATURE=2_4 BPM=74 TRACK_START INST=0 DENSITY=4 BAR_START NOTE_ON=52 TIME_DELTA=2.0 NOTE_OFF=52 NOTE_ON=45 NOTE_ON=49 TIME_DELTA=2.0 NOTE_OFF=49 NOTE_ON=52 TIME_DELTA=2.0 NOTE_OFF=45 NOTE_ON=47 NOTE_OFF=52 NOTE_ON=44 TIME_DELTA=2.0,
+  ...
+}
+```
+
+### Data Fields
+
+- `text`: Sequence of tokens that represent the guitar piece as explained in the paper [MMM: Exploring Conditional Multi-Track Music Generation with the Transformer](https://arxiv.org/abs/2008.06048).