nordicintel/common-crawl-se-discovery

Common Crawl Swedish URL discovery

Status: incomplete. 2,217,604 unique URLs and 15,832 unique hosts; 17,996 completed files in 60 selected crawls.

Frozen main crawls since 2020; .se hostnames, matching any term below, case-insensitively, with a non-alphanumeric boundary (or URL end) on each exposed alphanumeric edge. For example, ess. matches ess.example.se but not fitness.se; api/ matches /api/ but not /yogaterapi/. Punctuation stays literal; Unicode letters/numbers are alphanumeric and underscores are boundaries. URLs preserve their exact captured spelling and are globally deduplicated. Keywords identify candidates; their live availability, data content and licenses have not been verified.

[
  "/pxweb",
  "entryscape",
  "/catalog.",
  "/api",
  ".api",
  "api/",
  "opendata",
  "data.",
  "open-data",
  "arcgis",
  "gis.",
  "geodata",
  "map.",
  "karta.",
  "kartportal",
  "geoportal",
  "wms",
  "catalog.",
  "metadata.",
  "/metadata",
  "doi.",
  "ess.",
  "dataset",
  "openarchive",
  "oai",
  "geonode",
  "diva-portal",
  ".diva",
  "nvgis.",
  "kartor.",
  "geoserver",
  ".jsonld"
]

data/urls.parquet is a sorted single-column URL table. state/run.json reports coverage and failures; state/inventory.json freezes crawl numbering. state/checkpoint.sqlite3 resumes the same selection, skipping completed files and safely replaying partial files. All files advance together in one commit at each crawl boundary and graceful stop. Rerun the same harvest job after timeout.

Shortest URLs by host and pattern

derived/host-patterns.parquet has 15,832 rows: host followed by 32 pattern columns. Each non-null cell is the shortest full captured URL matching that pattern on that host, with binary lexical tie-breaking. Missing matches are null. This is a frozen derived table; derived/host-patterns.json records its source revision and matching rules.