pubmed-20k-rct / README.md
pietrolesci's picture
Upload README.md with huggingface_hub
d1741f4
|
raw
history blame
959 Bytes
metadata
task_categories:
  - text-classification
language:
  - en
dataset_info:
  features:
    - name: abstract_id
      dtype: string
    - name: labels
      dtype: string
    - name: text
      dtype: string
    - name: sentence_id
      dtype: int64
  splits:
    - name: train
      num_bytes: 33100479
      num_examples: 176642
    - name: validation
      num_bytes: 5552080
      num_examples: 29672
    - name: test
      num_bytes: 5516320
      num_examples: 29578
  download_size: 20080324
  dataset_size: 44168879

This is the same dataset as armanc/pubmed-rct20k. The only differences are

  1. Addition of a unique identifier, uid

  2. Addition of the indices, that is 3 columns with the embeddings of 3 different sentence-transformers

    • all-mpnet-base-v2
    • multi-qa-mpnet-base-dot-v1
    • all-MiniLM-L12-v2
  3. Renaming of the label column to labels for easier compatibility with the transformers library