# Source and derived-data notices

STO – SprogTeknologisk Ordbase / STO morphology v2, University of Copenhagen, Center for Language Technology (CST), 2008. Source record: http://hdl.handle.net/20.500.12115/21 . Copyright and the original embedded README are included byte-identically. The operative original grant is Creative Commons Attribution-ShareAlike2.5 Denmark: https://creativecommons.org/licenses/by-sa/2.5/dk/ . The archived license legal code is included. The repository's later4.0 catalog label is documented but not used to replace or infer the original grant.

Changes made for this audit: select only exact raw written-surface whole-form frequency fields for Korpus90/Korpus2000; collapse repeated identical analysis-row count pairs; sum the two corpus counts once; bind them to unchanged COR search keys; derive competition ranks, missing statuses, a deterministic domain and audit binary; compute coverage and inspect a general Phrase candidate floor. There is no lemma or POS-count propagation, case-alias merging or endorsement claim.

The STO-derived frequency binary, domain/count mappings and STO-specific frequency statistical extracts are offered under CC BY-SA2.5 Denmark. Raw large source archives/count tables are retained outside the review ZIP. The included STO documentation, README and copyright retain their original notices. This grant is compatible with commercial use subject to its conditions; it does not grant trademark or endorsement rights.

COR1 1.5.1.0, Dansk Sprognævn: CC0 1.0, https://creativecommons.org/publicdomain/zero/1.0/ . Original and derived dictionary/morphology data may be redistributed. Complete source rows are preserved for audit analysis, with unchanged v1 normalization and surface selection. The separate COR core is not relicensed by the STO sidecar.

UD Danish DDT r2.18 at commit36dc40e43302925941a60fb0201907ce50efbeb1: https://github.com/UniversalDependencies/UD_Danish-DDT/tree/r2.18 . Attribution: Anders Johannsen, Héctor Martínez Alonso, Barbara Plank; original Danish Dependency Treebank by Matthias Buch-Kromann, Line Mikkelsen and Stine Kern Lynge; PAROLE-DK / Danish Society for Language and Literature. The original README and CC BY-SA4.0 text are included.

Changes to UD material: preserve the v1 selected contiguous controls and source locations; derive relation-shape/direction counts and comparisons; record complete-row Danish projection/signals and control results. The UD-derived evidence and the combined control/quality traces containing both UD-derived annotations and STO-derived counts are distributed under CC BY-SA4.0. The STO original2.5DK grant permits a later license with the same license elements for adaptations; this combined-evidence choice does not relabel the original STO archive grant. No full treebank is packaged and no UD production-frequency sidecar is shipped.

Repository source references are unchanged user repository context for private review. These notices do not relicense application code. Rejected-source comparison reports contain factual identities, short license evidence and metadata; they do not grant redistribution rights to the rejected corpora. No full Leipzig corpus is included. No source authority endorses this audit or a future product.


# INS-110-PREQ-F2 source and derivative notices

## Danish Parliament binary adjacency

Source: The Danish Parliament Corpus 2009–2017, v2, with subject annotation (2021), Dorte Haltrup Hansen and Costanza Navarretta, Centre for Language Technology, NorS, University of Copenhagen. Original parliamentary material: The Danish Parliament / Folketinget.

Persistent record: http://hdl.handle.net/20.500.12115/44

Exact item: https://repository.clarin.dk/items/a3621457-73ec-412d-9d49-fb5db1e2d441

License: Creative Commons Attribution 4.0 International, https://creativecommons.org/licenses/by/4.0/ ; preserved legal text is included as INS-110-PREQ-F2-parliament-cc-by-4-legalcode.html. Retain attribution, source and license links, and indicate changes. Do not imply endorsement. The original README is included.

Changes in this audit: consume only the Text field from ten session archives; deterministically split sentence units and normalize exact Danish word surfaces; intersect a fixed Phrase candidate domain; discard counts and retain directed adjacent-word presence; encode sorted binary adjacency by word ordinal. No whole speech, sentence, speaker identity, demographic field or original frequency is included in the sidecar. The small audit witness table contains source coordinates and two-word facts only. The binary adjacency derivative and its witness table are distributed under CC BY 4.0. If raw speech text is redistributed separately, retain the source-required individual speaker attribution; this review does not redistribute raw speeches.

## Carried-forward dictionary, morphology, frequency and controls

COR1 1.5.1.0, Dansk Sprognævn: CC0. Complete-row excerpts, structural profiles and exact Danish normalization derive from this unchanged source.

STO morphology v2 / CST University of Copenhagen, whole-surface Korpus90 and Korpus2000 frequencies: original CC BY-SA 2.5 Denmark source grant. The frequency sidecar and domain remain byte-identical carried-forward derivatives; their original identity, attribution, changes, original license and ShareAlike obligations remain as stated in INS-110-PREQ-F1-NOTICES.md and alternative-frequency evidence. No Parliament counts enter those frequencies.

UD Danish DDT controls and syntax evidence: original attribution and CC BY-SA 4.0 obligations are preserved in the included UD README/license/ref and F1 notices. This source is not part of the binary adjacency corpus. Original corpus examples are used as carried-forward evaluation evidence, without claiming statistical independence from the source-backed grammatical calibration.

The combined audit runtime and quality tables contain adapted COR analyses and carried-forward STO frequencies plus UD-derived control selection. Distribute these data tables under CC BY-SA 4.0 with all component notices retained; the original STO source grant remains CC BY-SA 2.5 Denmark. The 2.5 license permits a later license version with the same elements for adaptations. Parliament binary data remain separately attributable CC BY 4.0. Audit code is not corpus content and repository code excerpts retain their existing repository status.

Danish Gigaword/TV2 Regionerne was inspected as a candidate but supplies no shipped adjacency/runtime data. Its exact candidate source and license distinctions are documented in sequence-report.md; no umbrella grant is inherited by the Parliament artifact.


# INS-110-A source and derivative notices

Source: The Danish Parliament Corpus 2009–2017, v2, with subject annotation (2021), Dorte Haltrup Hansen and Costanza Navarretta, Centre for Language Technology, NorS, University of Copenhagen. Original parliamentary material: The Danish Parliament / Folketinget.

Persistent record: http://hdl.handle.net/20.500.12115/44
Exact item: https://repository.clarin.dk/items/a3621457-73ec-412d-9d49-fb5db1e2d441
License: Creative Commons Attribution 4.0 International, https://creativecommons.org/licenses/by/4.0/ . Original README and legal text are included in the carried-forward evidence. Retain attribution, source/license links and change notices; no endorsement is implied.

INS-110-A changes: consume the Text field from all ten pinned session archives; apply unchanged F2 sentence/token normalization; count every exact directed adjacent pair whose tokens belong to the locked runtime domain; convert positive counts to global competition ranks; encode sorted ordinal/rank records and canonical source bindings. Missing pairs receive an explicit sentinel at lookup time. Sequence evidence changes ranking only and is not an eligibility gate. The binary and numeric occurrence/rank evidence are CC BY 4.0 derivatives. No raw speeches, speaker fields or download archives are included.

COR1 1.5.1.0 / Dansk Sprognævn remains CC0. The unchanged STO morphology v2/Korpus90/Korpus2000 exact-surface frequency derivative and its source attribution, original CC BY-SA 2.5 Denmark grant and adaptation obligations are preserved in INS-110-PREQ-F1-NOTICES.md. UD Danish DDT control examples retain original source attribution and CC BY-SA 4.0 notices. The combined runtime/control/result tables remain under CC BY-SA 4.0 as carried forward, with the original component notices. Parliament sequence numeric data are separately attributable CC BY 4.0. No source license is newly inferred or audited here.


INS-110-A-F1 adds a compact derived exact-trigram lookup and tier ranking to the same attributed Parliament source. No new corpus or rights decision. Tier data is a derived adaptation; original attribution and CC BY 4.0 notices above apply.


Production changes (INS-110-B): deployable COR dictionary and complete-row morphology; separately attributed exact Parliament numeric sequence data encoded by lossless ordinal remapping, column streams and gzip; frozen ranking and primary/fallback partition. No endorsement is implied. COR dictionary remains CC0; combined morphology/frequency derivative CC BY-SA 4.0; Parliament sequence derivative CC BY 4.0. Original source grants remain unchanged.