Update README.md

README.md

@@ -48,7 +48,7 @@ The vocabulary is then used to filter the widely used text corpora,
 - Simplified-Wikipedia: https://simple.wikipedia.org/wiki/Main_Page,
 - Children's Book Test Corpus: https://arxiv.org/abs/1511.02301
 
-From the above corpora, only those spans are included that contain words only from the predefined vocabulary. The dataset includes 44 million sentences (~6 million sequences of ~128 tokens) and 3 million contiguous spans with approximately 128 tokens.
+From the above corpora, only those spans are included that contain words only from the predefined vocabulary. The dataset includes 44 million sentences (~6 million sequences of ~128 tokens) and 3 million contiguous spans with approximately 128 tokens. Refer to Table 1 of the paper for data distribution over different corpora.
 
 ### Languages
 
@@ -56,11 +56,7 @@ The dataset contains the English language only.
 
 ## Dataset Structure
 
-The dataset is available in the arrow dataset format with three splits namely, train, validation and test.
-
-### Data Instances
-
-Every data instance has only one key "Text" that included a text span of approximately 128 tokens.
+The dataset is available in the arrow dataset format with three splits namely, train, validation, and test. Every data instance has only one key "Text" that included a text span of approximately 128 tokens.
 
 ### Citation Information