metadata

license: apache-2.0
dataset_info:
  features:
    - name: coreId
      dtype: string
    - name: title
      dtype: string
    - name: authors
      sequence: string
    - name: datePublished
      dtype: string
    - name: abstract
      dtype: string
    - name: relations
      sequence: string
    - name: year
      dtype: int64
  splits:
    - name: core_2018
      num_bytes: 113101903827
      num_examples: 85610399
  download_size: 64859628168
  dataset_size: 113101903827
configs:
  - config_name: default
    data_files:
      - split: core_2018
        path: data/core_2018-*
task_categories:
  - question-answering
  - summarization
  - text2text-generation
  - sentence-similarity
language:
  - en
  - zh
tags:
  - chemistry
  - biology
  - legal
  - finance
  - music
  - art
  - climate

A greek scholar writing manuscript in 1080BC

COREX 18

Introducing COREX-18, a comprehensive dataset derived from the 2018 version of the CORE dataset. Our goal is to contribute to the research community by compiling open-access scientific papers and publishing them in extensive datasets. These datasets will facilitate advanced RAG applications and enhance artificial intelligence research.

COREX was developed as part of our X initiative, which aims to maintain and compile publicly available data into accessible and regularly updated datasets.

COREX-18 consists of over 85 million rows.

Regarding metadata preservation, we made the decision not to include all metadata due to its complexity and the high number of NULL values encountered. Instead, we preserved critical metadata essential for tracking research papers and understanding their basic information.

COREX-18 remains unaltered from the original CORE dataset. We maintained the integrity of the abstracts and titles without subjecting them to any textual cleaning processes.

COREX-18 is primarily targeted towards RAG Applications and the Citing data and scientific knowledge category.

Note: Full-text version will be updated soon.