Update README.md

README.md

@@ -3,7 +3,9 @@ language:
 - en
 tags:
 - chemistry
-dataset_info: 
+- pubchem
+- wikipedia
+dataset_info:
   features:
   - name: sent1
     dtype: string
@@ -22,8 +24,14 @@ configs:
   data_files:
   - split: test
     path: data/test-*
+license: cc-by-nc-sa-4.0
+task_categories:
+- text-classification
+pretty_name: PubChem & Wikipedia Paragraphs Pair Classification
+size_categories:
+- 1K<n<10K
 ---
 
 # PubChemWikiParagraphsPC
 
-This PubChem AI pair classification task comprises of some pubchem entity descriptions. Negative pairs are paragraphs from different sources of PubChem, and positive pairs are paragraphs of the same entity across two different sources.
+This dataset is for Pair Classification and has three columns: `sent1`, `sent2`, and a binary column `labels`. The `labels` is 1 when `sent1` and `sent2` describe the same entity and 0 when they describe two different entities. The dataset consists of 2048 pairs sourced from [PubChem](https://pubchem.ncbi.nlm.nih.gov/) ([LOTUS](https://lotus.nprod.net/) and [ChEBI](https://www.ebi.ac.uk/chebi/) sources) and English Wikipedia (article summary) with a 90-10% negative-positive ratio.