BASF-AI/PubChemWikiParagraphsPC

---
language:
- en
license: cc-by-nc-sa-4.0
size_categories:
- 1K<n<10K
task_categories:
- text-classification
pretty_name: PubChem & Wikipedia Paragraphs Pair Classification
tags:
- chemistry
- pubchem
- wikipedia
- chemteb
dataset_info:
features:
- name: sent1
dtype: string
- name: sent2
dtype: string
- name: labels
dtype: int64
splits:
- name: test
num_bytes: 1407915
num_examples: 2048
download_size: 739062
dataset_size: 1407915
configs:
- config_name: default
data_files:
- split: test
path: data/test-*
---

# PubChem & Wikipedia Paragraphs Pair Classification

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.