Sources

This page credits the sources used in Apera's word lists, coverage figures, maps, and cultural catalogues. Each section names the source and links to its licence or publisher, and the last but one says how the film, book and song lists are made.

The order the Speaking lists are in

The Speaking lists — the Italian, French, German and Spanish lists of the words people say — are sorted by a published table of word counts — the 2018 release of Hermit Dave’s FrequencyWords, taken at a pinned commit. Its author licenses the contents under CC BY-SA 3.0. Every word on the Speaking lists is a word from that table, kept in its order; for these lists we count nothing ourselves.

The Speaking lists themselves — the files each Speaking page offers — are published under CC BY-SA 4.0, and each file ends with a line naming that licence and linking here. You may reuse them, including commercially, on that licence’s terms — among them, crediting the sources on this page and sharing anything you build from them under the same licence.

Where those words were counted

The counts behind that table — and which words are on the Speaking lists at all — come from subtitles: the OpenSubtitles collection as distributed by the OPUS project at the University of Helsinki. The subtitles are written by volunteers rather than by the programmes they transcribe, which is why the counts describe how people speak on screen rather than how anyone writes.

OPUS distributes the collection under the Open Data Commons Attribution License (ODC-BY), which covers its own work of compiling and aligning it — not the subtitle text itself, which OPUS does not claim to own. For the 2018 release its maintainers ask for a citation of P. Lison and J. Tiedemann (2016), “OpenSubtitles2016: Extracting Large Parallel Corpora from Movie and TV Subtitles”, and a link to opensubtitles.org. We honour both.

The Reading lists

The Reading lists — the most common words of writing about one topic, such as news or sport — are the reverse case: we counted them, from pages on sites that had not opted out of text and data mining when their pages were read. Nothing from any page is published but single words in the order of their counts.

A Reading page compares its words with the language’s Speaking list, showing each word’s Speaking rank beside its rank in the topic’s writing. Those ranks are only used for that comparison, never in the files; they come from FrequencyWords, credited above under its licence, CC BY-SA 3.0.

Each list is counted across all of its topic’s sites together, never as one list per site. No count of any word is published, nothing is given per site, and a word only one site used is left out.

• Italian news vocabulary — counted from 687 sites
• Italian sports vocabulary — counted from 148 sites
• Italian food vocabulary — counted from 121 sites
• Italian tech vocabulary — counted from 75 sites
• French news vocabulary — counted from 239 sites
• French sports vocabulary — counted from 162 sites
• French food vocabulary — counted from 111 sites
• French tech vocabulary — counted from 110 sites
• German news vocabulary — counted from 218 sites
• German sports vocabulary — counted from 162 sites
• German food vocabulary — counted from 219 sites
• German tech vocabulary — counted from 194 sites
• Spanish news vocabulary — counted from 245 sites
• Spanish sports vocabulary — counted from 131 sites
• Spanish food vocabulary — counted from 107 sites
• Spanish tech vocabulary — counted from 76 sites

The files are published under CC BY-SA 4.0 on the same terms as the Speaking files above, and each ends with a line naming that licence and linking here. The English meanings are short guides that can change with context, as on the Speaking lists.

The coverage figures

Where a word list is described as covering some share of ordinary writing, and where a story on the placement page has to place a word rarer than our own list reaches, those figures are read from wordfreq rather than from our own ranking — deliberately, so that a figure describing how far our list gets you is not measured with the list itself. The word shading you see after the quiz reads our own list, not wordfreq. The wordfreq package is released under the Apache License 2.0, and the data it ships is published under the Creative Commons Attribution-ShareAlike 4.0 licence.

The exporter sums wordfreq frequencies for the forms in each downloaded list. That produces an estimate of word occurrences in the reference data, not a measurement of any particular page or of how much a reader understands.

The relief maps on the region pages

The shaded land behind each region map is generated once from public elevation data and served as a picture; nothing is fetched while you read it. The heights over Italy come from EU-DEM, produced using Copernicus data and information funded by the European Union, reached through the Terrain Tiles dataset on the Registry of Open Data on AWS, which also contributes ETOPO1 from the U.S. National Oceanic and Atmospheric Administration over the sea. The relief is derived from those elevations rather than reproduced from them — we have modified the data — and it is not endorsed by the European Union. The region outlines and the small shapes on the index are Natural Earth, which asks for no credit and is given one anyway.

The relief plates on the Spanish-language film region pages

The relief behind each Spanish-language film region page is generated once from public elevation data and served as a picture, the same way as the Italian region maps above, from a different mosaic: heights over Spain come from EU-DEM, produced using Copernicus data and information funded by the European Union; heights over the Americas come from the Shuttle Radar Topography Mission (SRTM), GMTED2010 and ETOPO1, all from U.S. federal sources. The relief is derived from those elevations rather than reproduced from them — we have modified the data — and none of it is endorsed by the European Union. The region outlines are Natural Earth, which asks for no credit and is given one anyway.

The relief plates on the French-language film region pages

The relief behind each French-language film region page is generated once from public elevation data and served as a picture, the same way as the maps above, from a third mosaic covering metropolitan France: heights come from EU-DEM, produced using Copernicus data and information funded by the European Union, with the Shuttle Radar Topography Mission (SRTM), GMTED2010 and ETOPO1 — all from U.S. federal sources — where EU-DEM does not reach. The relief is derived from those elevations rather than reproduced from them — we have modified the data — and none of it is endorsed by the European Union. The region outlines are Natural Earth, which asks for no credit and is given one anyway.

The relief plates on the German film and book region pages

The relief behind each German film and book region page is generated once from public elevation data and served as a picture, the same way as the maps above, from a mosaic covering Germany: heights come from EU-DEM, produced using Copernicus data and information funded by the European Union, with the Shuttle Radar Topography Mission (SRTM), GMTED2010 and ETOPO1 — all from U.S. federal sources — where EU-DEM does not reach. The relief is derived from those elevations rather than reproduced from them — we have modified the data — and none of it is endorsed by the European Union. The region outlines are Natural Earth, which asks for no credit and is given one anyway.

Who wrote a book

A book’s entry names its author, and that name comes from Wikidata — the same CC0 record the title, the year and the identifiers on that row come from, published under a waiver that asks for no credit. A row lists at most three names and then says and others. Where a row names nobody, Wikidata records no author for the work: that is a gap in the catalogue rather than a work published anonymously, and the section on the lists below says how many of each book list’s titles name one. We publish nothing else about the person — no dates, no nationality, no link.

Where a title is set, and what form a book takes

The region pages say a title — a film, a series or a book — is set somewhere on three authorities. The first is Wikidata, whose statement of where a work is set is published CC0 and asks for nothing. The other two are Italian Wikipedia and English Wikipedia, whose editors file a title under categories like Film ambientati a Roma and Television shows set in Naples — set there, which is a different claim from filmed there, and only the first puts a title on one of these pages. Those categories are part of each encyclopedia’s content and are published under the Creative Commons Attribution-ShareAlike 4.0 licence. We read the category tags and the tree above them, from each Wikipedia’s own published database exports, and keep no article text.

Italian Wikipedia’s categories answer a second question about a book: whether it is prose or verse and drama, filed under categories like Romanzi and Poemi. We read that only where Wikidata’s own record of a work leaves the question open, and never to override what Wikidata already says. Same licence, same condition, same credit.

The film, book and song lists

Titles, years, languages and authors come from Wikidata (CC0). Lookup figures come from Wikimedia pageviews. Settings on the Italian region pages also use Italian and English Wikipedia categories, under CC BY-SA 4.0 (credited above). We hold no text of these works and do not measure their difficulty.

Ranked lists use average daily visits to a title’s article on the list’s own Wikipedia, August 2021–July 2026. These measure lookups — not viewings, tickets, reading, sales, library loans, streams, plays, airplay or chart position. School reading lists, anniversaries and controversies can move the order.

“Only on” means exactly one Wikipedia language edition, the list’s own, has an article about the title. That decides who is listed, never the order. It establishes neither quality nor obscurity nor unavailability elsewhere. Titles with no Wikipedia article in any language are not on these lists.

Films and series

A row’s stated languages reproduce Wikidata’s original-language records, named in the list’s language and in alphabetical order: Wikidata gives them no priority. They do not identify the audio of any particular version.

• Italian: titles tagged erotic, pornographic or commedia-sexy are excluded; extreme cinema is kept. The filter is known to miss some titles.
• Full manual review pending: titles may be missing or wrongly included.

Only on Italian Wikipedia: 42.4% of the films are from after 2015, against 7.0% of films with an article in eight or more Wikipedia editions, so part of what keeps a film to one Wikipedia is age. The series show no such pattern: 16.8% are from after 2015.

Books

A ranked list needs an article on the list’s own Wikipedia averaging at least one visit a day over the window. Only-on lists have no lookup minimum; a row with no recorded lookups says so rather than printing a zero. Comic serials are excluded by class; graphic novels stay. Author names come from Wikidata, and a row naming nobody means Wikidata records no author — a gap in the record, not anonymity.

• French: any variety of French counts — Quebec, Belgium, Switzerland, Haiti and francophone Africa as well as France.
• German: any variety of German counts — Austria, Switzerland, Liechtenstein, Luxembourg and German-speaking Belgium as well as Germany.
• French: 1,689 titles that would otherwise be listed are excluded as comics; 125 graphic novels are kept.
• German: 3,573 catalogue records are excluded as comics; 13 graphic novels are kept.
• Literary form (prose, or poem or play) is unrecorded for 80% of ranked French titles, 81% of ranked Spanish titles and 75% of ranked German titles, so those lists have no form filter.
• Spanish: Fundación SCP is filed by Wikidata as a literary work but is not a book; it is kept and flagged rather than removed.
• French, Spanish and German: classification, including an adult-content review, is unfinished.

Songs

A song needs an article on the list’s own Wikipedia; for the Italian list that is the whole rule. The others also need one of these:

• French: Wikidata records French, or a variety of it, as the song’s language, or names a performer or composer documented as French-speaking.
• Spanish: Wikidata records Spanish as the song’s language, or names a performer or composer documented as Spanish-speaking.
• German: Wikidata records German, or a variety of it, as the song’s language, or names a performer or composer holding citizenship of a country where German is an official language.

• Italian: 33 of the records failing the criteria are art music — masses, requiems, sonatas — that Wikidata files under the same class as a single.
• Le Chant des partisans is on the French list by a named exception rather than the ordinary rule: Wikidata states the language of Anna Marly's Russian original; the song known by this title is the French Resistance anthem.
• German: most records not listed are compositions, which Wikidata catalogues by composer where a release carries a performer.

Settings and maps

Maps place a title by where its story is set, not where it was filmed, written or published. A title set in two regions counts in both. Italian film maps leave out the 80 titles placed only by a filming location. French, Spanish and German film pages list such titles apart, labelled, and never add them into a region’s figures.

Titles set outside the country sit on the last row of a section’s table, not on its map.

• Italian books: Umbria holds nothing and keeps its page and map.
• Spanish books: La Rioja holds nothing and keeps its page and map.
• German films and series: Saarland holds nothing and keeps its page and map.
• German films and series: Rheinland-Pfalz holds fewer than three titles and is drawn all the same.
• German books: Bremen, Saarland and Mecklenburg-Vorpommern hold fewer than three books and are drawn all the same.

If something here is wrong

These are claims about other people’s work, and we would rather be corrected than right by default. Write to hello@apera.me and say which entry and what is wrong with it.

The film pages in particular show what we can see in the data. The records have gaps and the occasional plain error, and every description of how a book reads is written by an AI model, Claude Opus, from what it knows of the title: a judgement rather than a measurement, and nobody has checked those descriptions one by one. So if a page disagrees with what you have heard, believe your ears. The best correction to any of it is a trip, and a question to somebody who lives there; the second best is the address above.