Leipzig Corpora Collection — Universität Leipzig / Sächsische Akademie der Wissenschaften / InfAI. CC BY; version not stated by upstream. No CC BY version is inferred. # Leipzig Corpora Collection Italian frequency notice The derived Italian Phrase frequency sidecar uses only the officially downloadable `ita_news_2024_1M` archive and its exact `ita_news_2024_1M/ita_news_2024_1M-words.txt` member. Source attribution: Leipzig Corpora Collection; Universität Leipzig; Sächsische Akademie der Wissenschaften; InfAI. Project citation: Dirk Goldhahn, Thomas Eckart and Uwe Quasthoff, “Building Large Monolingual Dictionaries at the Leipzig Corpora Collection: From 100 to 200 Languages”, LREC 2012, pages 759–765, https://aclanthology.org/L12-1154/. Instant Anagram verifies the pinned archive and members, decodes each source row as strict UTF-8, maps whole word forms with NFC → Italian locale lowercase → NFC, sums only distinct raw rows that map to the same exact normalized surface, and derives zero-based competition ranks. It does not accent-fold, stem, substitute lemmas, borrow morphology, rescale counts, or infer a corpus denominator. Missing source evidence remains explicit `UNRANKED` (`0xffffff`) and does not remove a candidate. The values are observed surface-token counts from an Italian news corpus. They are not lemma frequencies, phrase probabilities, grammaticality judgments, semantic plausibility scores, or general-language absolute-frequency estimates. No endorsement is implied and no additional restriction is applied. # Leipzig Italian news frequency source ## Locked authority - Source id: `it-leipzig-ita-news-2024-1m-frequency` - Name: Leipzig Corpora Collection Italian news normalized exact-surface frequencies - Corpus/version: `ita_news_2024_1M / build 2025-02-12 / toolchain 406a2d7824b1a2b4b2f4b79be729fd2b7b8296ec / resources 681b682caa6c3c8250f8ab2c51ba3acc1c579959` - Description: https://wortschatz.uni-leipzig.de/en/download/ita - Official download: https://downloads.wortschatz-leipzig.de/corpora/ita_news_2024_1M.tar.gz - Archive: `ita_news_2024_1M.tar.gz`, 261325052 bytes, SHA-256 `90ed9f839c50de9a58a7f1772516891d6721c92f4b863560bd26a480d7717e30` - Frequency authority member: `ita_news_2024_1M/ita_news_2024_1M-words.txt`, 9027017 bytes, SHA-256 `9246734e719c87a73f1fc93c7237040e4a7f38337e99d55d5ff5b181a0a61d67` - Identity metadata member: `ita_news_2024_1M/ita_news_2024_1M-meta.txt`, 338 bytes, SHA-256 `c1c5d1664d8f2e65d1dd16b5b22b06d4bba4cd99619746285b841907d459aef6` - Source format: https://wortschatz.uni-leipzig.de/public/documents/Format_Download_File-eng.pdf (`Word_ID`, `Word`, `Frequency`, tab-separated UTF-8) - Terms: https://wortschatz.uni-leipzig.de/en/usage - License: `CC BY; version not stated by upstream` - Attribution: Leipzig Corpora Collection; Universität Leipzig; Sächsische Akademie der Wissenschaften; InfAI The upstream downloadable-corpus grant states: “The text corpora offered for download are made available under the Creative Commons licence CC BY.” The license version is not inferred. Preserve the grant, terms locator, attribution, identification of changes, no-endorsement condition and absence of additional restrictions. Project citation: Dirk Goldhahn, Thomas Eckart and Uwe Quasthoff, “Building Large Monolingual Dictionaries at the Leipzig Corpora Collection: From 100 to 200 Languages”, LREC 2012, pages 759–765, https://aclanthology.org/L12-1154/. Citation guidance: https://wortschatz-leipzig.de/en/publications. ## External reproduction boundary The deterministic boundary begins with the exact pinned archive outside the repository. Generation verifies the archive byte length and SHA-256 before reading the exact `_words.txt` and `_meta.txt` members. Only the word member's raw `Word` and `Frequency` fields supply frequency authority. The metadata member verifies corpus identity, build date, inventory and pinned toolchain and resource revisions. The raw archive and extracted members are not tracked. Run from the repository root with the pinned archive and exact committed input packages, writing to a new, empty directory outside the repository: ```text tsx tools/generate-it-frequency-sidecar.ts PATH_TO/ita_news_2024_1M.tar.gz PATH_TO/it-core.bin PATH_TO/it-phrase-short-words.json NEW_EMPTY_OUTPUT_DIRECTORY ``` The generator writes `it-frequency..bin`, `manifest.json` and `frequency-coverage.json`. Compare all three outputs byte-for-byte with the committed foundation. The metadata reports `WORD_TOKENS=22273679`; the authoritative word-row frequency sum is 22607384. The 333705 difference is retained as a source limitation. Counts are not rescaled and no denominator is inferred. ## Instant Anagram transformations 1. Decode the word list as strict UTF-8 and retain each distinct raw row. 2. Normalize a whole source `Word` with NFC → `toLocaleLowerCase("it")` → NFC. 3. Sum only distinct raw rows that produce the same exact normalized surface. Do not accent-fold, stem, split, substitute a lemma, borrow morphology or use fuzzy matching. 4. Match only normalized surfaces in the exact pinned Italian Phrase core plus one-letter overlay domain and retain their observed raw counts. 5. Rank positive surfaces by the number of positive domain surfaces with a strictly greater count. Equal counts share a zero-based competition rank. 6. Encode missing evidence as `UNRANKED` (`0xffffff`), without imputing a count or changing candidate eligibility. The immutable ITF1 foundation retains inactive foundation metadata (`phraseEnabled:false` and `activeRuntimePackage:false`). The active `it-realizability-v1` profile uses it through the outer runtime descriptor. This Leipzig package is a public source-reproduction and attribution package. Bundled historical foundation documentation retains its original pre-activation context. The documented 333 705 token-sum discrepancy is preserved. Counts are not rescaled. The license applies to Leipzig-derived frequency data; external core and short-word inputs retain their separate Morph-it/Wikizionario ShareAlike attribution chain in the Italian Phrase source bundle.