Search, rank and compare how films actually speak. The first live corpus focuses on reported F-word usage; the platform is built to expand into scene-level timestamps, speakers, actors, cuts, censorship and broader language categories without inventing data that is not sourced.
Search the corpus, filter by decade and computed density, bookmark locally, compare up to four titles, and attach a supported YouTube or Vimeo trailer to any record. The app attempts a live source refresh while retaining a dated snapshot for resilience.
Reported totals come from the source corpus. Density is independently computed as reported count ÷ runtime, so a shorter film can outrank a longer one even with fewer total uses.
Select up to four films and compare total usage, rate per minute and runtime. The comparison state can be copied into the URL for sharing.
Select films to compare.
The data model can extend beyond aggregate counts. A sourced language event can carry a timestamp, speaker, actor, context, cut, censorship behavior, confidence and provenance. Unknown fields remain null instead of being guessed.
{
"film_id": "goodfellas-1990",
"word": "fuck",
"timestamp_seconds": null,
"speaker": null,
"actor": null,
"context": null,
"cut": null,
"censorship": null,
"confidence": null,
"source_url": null
}Definitions explain what the numbers do and do not mean. Quizlets reinforce the difference between reported source fields, computed metrics and future scene-level evidence.
Cinema Swear Stats separates source-reported facts, independently derived metrics and future enrichment. The initial F-word corpus inherits the limitations of its source, so citation quality and math discrepancies remain visible.
The first corpus is based on Wikipedia’s dynamic list of English-language feature films reported to contain at least 150 uses of the F-word or derivatives. Snapshot captured 2026-11-09. The source itself warns about original research and mixed source reliability.
Computed rate is count ÷ runtime and powers density rankings. Source rate preserves the value displayed by Wikipedia. If the two differ, the record is flagged instead of silently choosing one.
The browser attempts to parse the current source table through MediaWiki. If that fails, the 2026-11-09 snapshot remains usable. Static film pages intentionally represent that dated snapshot.
Timestamp, speaker, actor, censorship and alternate-cut fields will only render when a traceable source is stored. Unknown fields remain null.
No. The first corpus visualizes a dynamic source with uneven reference quality.
No. Rating decisions involve context and exceptions; this app does not infer ratings from counts alone.
Long films can accumulate large totals. Cinema Swear Stats computes count ÷ runtime so density can be compared independently from the source-displayed rate.
Yes, when sourced corpora can be added without fabricating counts or scene metadata.
Download the dated snapshot, the schema that validates it, or export the currently loaded live view directly from your browser. Static film pages provide indexable records for the captured corpus.
Snapshot files are dated 2026-11-09. “Current” exports use whichever dataset is actually loaded in your browser: live source when sync succeeds, otherwise the packaged snapshot.