ffatty/plaintext-wikipedia-full-norwegian

Plaintext Wikipedia (Norwegian)

Compressed plaintext articles from Norwegian Wikipedia. One current main-namespace revision per non-redirect article; titles and cleaned text are stored as UTF-8 in independent gzip shards.

Produced and maintained by cheeseonamonkey/PlainTextWikipedia.

Dataset availability

2 compressed archives currently published. Source recovery: 2/2 roots, status COMPLETE. See data/build-manifest.json for exact missing or incomplete ranges.

Format

Each .txt.gz shard has three lines per article (title, single-line body, blank separator):

Article title
Cleaned article body

Next title
Next article body

Shard files are normally named nowiki_01_of_NN.txt.gz, etc. If one shard fails, recovery splits it into named halves like nowiki_01a_of_NN.txt.gz and nowiki_01b_of_NN.txt.gz; a failing half can split again (01ba, 01bb, etc.). Only successful, non-overlapping leaves are published.

The machine-readable coverage file data/build-manifest.json records which roots and subranges completed. Always check its complete flag before treating a dataset as exhaustive. These are raw text archives, not structured rows with title and body fields in the Hugging Face datasets API.

Structured editions (when available)

The structured/ directory contains independently verified Parquet derived from the original data; its presence varies by edition. Existing plaintext archives are retained. Older SQLite outputs may still be present from previous versions but are no longer generated.

File	Use	Format
structured/parquet/*.parquet	Fast tabular analysis, pandas, DuckDB, Polars	Parquet (Zstandard)
structured/manifests/*.parquet.json	Article count, raw SHA256, output SHA256	JSON

Parquet columns: title (string), text (string), source_file (string), source_ordinal (int64; 1-based within its source archive). Original Wikipedia page IDs are not retained.

from huggingface_hub import HfApi, hf_hub_download
import pyarrow.parquet as pq

repo = "ffatty/plaintext-wikipedia-full-norwegian"
files = sorted(path for path in HfApi().list_repo_files(repo, repo_type="dataset")
               if path.startswith("structured/parquet/") and path.endswith(".parquet"))
if not files:
    raise RuntimeError("No Parquet editions published yet")
path = hf_hub_download(repo, filename=files[0], repo_type="dataset")
print(pq.read_table(path, columns=["title", "text"]).slice(0, 3))

Coverage warning: verify data/build-manifest.json before describing an edition as complete; a populated archive collection does not itself prove full source coverage.

Example: download and read

import gzip
from huggingface_hub import HfApi, hf_hub_download

repo = "ffatty/plaintext-wikipedia-full-norwegian"
files = sorted(
    path for path in HfApi().list_repo_files(repo, repo_type="dataset")
    if path.startswith("data/") and path.endswith(".txt.gz")
)
path = hf_hub_download(repo, filename=files[0], repo_type="dataset")

with gzip.open(path, "rt", encoding="utf-8") as stream:
    title = stream.readline().rstrip("\n")
    body = stream.readline().rstrip("\n")
    print(title, body[:200])

Sources and attribution

1. cheeseonamonkey/PlainTextWikipedia — maintained converter and publisher.
2. daveshap/PlainTextWikipedia — original upstream project (unmaintained).
3. Wikimedia dumps — article source.

Wikipedia content is released under Creative Commons Attribution-ShareAlike 4.0 (CC BY-SA 4.0); the conversion code is MIT licensed. Follow the applicable Wikimedia reuse requirements, including author attribution.

---

Generated by cheeseonamonkey/PlainTextWikipedia. Source data: Wikimedia dumps.