dvc.lock

schema: '2.0'
stages:
  prepare-s2orc:
    cmd: python scripts/s2orc/prepare_parquet_s2orc.py
    deps:
    - path: mars/s2orc/metadata/ai/
      md5: 1bff989bf79e7b8cedccb963caf2c284.dir
      size: 193936339
      nfiles: 100
    - path: scripts/s2orc/prepare_parquet_s2orc.py
      md5: 1ec9038eda1aea162af1541aa944adf6
      size: 645
    outs:
    - path: data/s2orc/s2orc_ai_prefiltered.parquet
      md5: b7e11eba387eb511989bf74e37f9f950
      size: 73253586
  process-spacy-s2orc:
    cmd: python scripts/s2orc/process_spacy_docs.py
    deps:
    - path: data/s2orc/s2orc_ai_prefiltered.csv
      md5: 372b0a56816c5ecd1086344bea24498b
      size: 90660899
    - path: scripts/s2orc/process_spacy_docs.py
      md5: 58d88a3a785fd99d45feb05258bfd82f
      size: 509
    outs:
    - path: data/s2orc/s2orc_ai_prefiltered_processed_with_doi.pkl
      md5: ebfa849796b26e87f519e3372d811bad
      size: 5558460171
  extract-keywords-s2orc:
    cmd: python scripts/s2orc/extract_keywords.py
    deps:
    - path: data/s2orc/s2orc_ai_prefiltered_processed_with_doi.pkl
      md5: ebfa849796b26e87f519e3372d811bad
      size: 5558460171
    - path: scripts/s2orc/extract_keywords.py
      md5: 525a6a84dd496d9107cca48c2f7ee473
      size: 791
    outs:
    - path: data/s2orc/extracted.csv
      md5: bbbd22b86b56f7ca385f8326b52821e4
      size: 124316244