Datasets:

pymlex
/

future-engineers-dataset

dataset_info:
  features:
    - name: id
      dtype: string
    - name: year
      dtype: int64
    - name: topic
      dtype: string
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 8775081
      num_examples: 1976
    - name: validation
      num_bytes: 382317
      num_examples: 106
    - name: test
      num_bytes: 498108
      num_examples: 111
  download_size: 4320384
  dataset_size: 9655506
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
license: gpl-3.0
task_categories:
  - text-generation
language:
  - ru
pretty_name: Future engineers
size_categories:
  - 1K<n<10K

Provenance

This dataset is based on 2,200 project descriptions from the Future Engineers conference. The source PDFs are freely distributed on the official site.

Data processing

These files were stripped of images and then converted to Markdown using the datalab/marker-pdf utility. The raw MD files were manually edited, which included:

Personal data removal
Header formatting

Finally, the data was converted to JSONL format.