wikipar / README.md
dchaplinsky's picture
Initial upload of Wikimedia paragraphs dataset
adc1b80
# My Wikimedia Dataset
This dataset contains paragraphs extracted from Wikimedia Wikipedia texts. Each entry includes:
- `paragraph_text`: Text of the paragraph
- `original_url`: URL of the source article
- `original_id`: ID of the source article
- `paragraph_number`: Paragraph number within the original text
- `combined_id`: Unique hash ID for the paragraph
- `length`: Paragraph length in characters
- `length_tokens`: Paragraph length in tokens (Gemma tokenizer)
Format: Parquet (zstd compression).