README.md

metadata

language:
  - en
license: unknown
tags:
  - text
  - corpus
  - literature
  - project-gutenberg
  - pg19
pretty_name: PG-19 Filtered Enriched
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/*.jsonl

PG-19 Filtered Enriched

This dataset contains the PG-19 portion of the MonumentalSystems text pipeline, exported as cleaned JSONL shards.

What is included

• 498 JSONL shard files under data/
• 299,410 chunk rows total
• the original filtered PG-19 portion
• the continued 0364-0573 PG-19 tail processed through the same cleaner, chunker, quality scorer, and DataTrove filtering path

Row fields

Each row includes the original chunk text plus pipeline metadata such as:

• source coordinates (source_ref, file_index, doc_index, chunk_index)
• dedup fingerprints (checksum, normalized_checksum, alnum_checksum)
• quality and curriculum fields (reading_level, complexity, difficulty, flesch, mtld, quality_pass, quality_reason)
• semantic labels (content_type, subject, subject_detail, title_hint)

Notes

• This export is PG-19-only, not the full text corpus.
• The continuation portion was exact-deduped against the existing PG-19 filtered corpus before export.
• title_hint, subject, and subject_detail are heuristic enrichment fields and should be treated as best-effort metadata rather than bibliographic ground truth.