Planning inventory · no public downloads yet
A pack is a versioned artifact CleverMemory can read: a snapshot of public material, frozen at a known state, with its terms attached. The planned public packs cover language and general knowledge that can be distributed with clear source terms. Private product data stays private.
This page is the future download and attribution record. An entry in the inventory means a source is under review; it does not mean a usable pack has been released, and it says nothing about any other source or about the product itself.
How to read the status#
Each card carries a posture line saying where that source stands in review:
| Posture on a card | Meaning |
|---|---|
| Public-distribution candidate | Under review for broad reuse. Release checks are still open. |
| Subset review required | Only individually reviewed members of the source family may qualify. |
| Counsel confirmation before release | The grant reads as permissive; a final cross-check is still pending. |
| Tracked candidate | Being followed, but not yet pinned or scoped for a pack. |
| Public download planned | A ShareAlike source intended to publish as its own separate download. |
A source moves from candidate to released only when a versioned artifact and its manifest are available to download. In between, a candidate artifact may pass the published format checks in a declared test environment; that is controlled validation, and it is not release approval. No source should be treated as released without a public artifact and release manifest.
Source boundaries#
The inventory keeps three groups separate:
- sources being reviewed for broad public reuse;
- sources that must retain their own ShareAlike or similar terms; and
- the CleverMemory layer itself, the code and data that tie everything together, which will not be downloadable.
The rule behind the grouping is decided. A source enters the inventory only under public-domain, CC0, attribution, permissive, or ShareAlike terms; non-commercial terms are not admitted in any tier. Released packs will be hosted on this site with full attribution to the original datasets, and ShareAlike packs will always be separate downloads.
“Friendly” is a planning label, not legal advice. Each released pack will name its exact source, snapshot, terms, attribution, changes, and exclusions. A source-data description below says what the upstream material contains; it does not describe how CleverMemory builds or uses a pack.
Boundary 01
Friendly source candidates
Sources currently eligible for broad-reuse review because the selected material is public domain, CC0, attribution licensed, or permissively licensed. Each cleared pack is intended for public hosting here.
Wikidata entities and Lexemes*
Public-distribution candidateSelected materialRolling entity and Lexeme dumps
Source dataRecords community-contributed entity statements, labels, aliases, and qualifiers alongside lexical entries, forms, and senses.
TermsCC0 1.0
Open English WordNet*
Public-distribution candidateSelected material2024 edition
Source dataContains English synsets, word senses, definitions, usage examples, and semantic relationships between lexical concepts.
TermsCC BY 4.0
Open Multilingual Wordnet*
Subset review requiredSelected material1.4, permissive allowlist only
Source dataContains wordnet-style lexical concepts, word senses, labels, and semantic relationships across multiple languages.
Swedish FrameNet*
Public-distribution candidateSelected materialSweFN 2.0
Source dataAnnotates Swedish lexical units and sentence examples with semantic frames, frame elements, and participant roles.
TermsCC BY 4.0
PropBank Frames*
Public-distribution candidateSelected materialv2.1.5
Source dataLists predicate senses as rolesets with numbered argument roles, role descriptions, and annotated usage examples.
TermsCC0 1.0
Selected material3.4 from the cu-clear repository
Source dataLists English verbs by class and records thematic roles, syntactic frames, and semantic predicates for each class.
TermsHistorical VerbNet 3.X permissive grant; coverage under review
Universal Dependencies Spanish AnCora*
Public-distribution candidateSelected materialr2.16
Source dataAnnotates Spanish sentences with tokens, lemmas, parts of speech, morphological features, and dependency relations.
TermsCC BY 4.0
ATOMIC 2020*
Public-distribution candidateSelected materialFebruary 2021 train, development, and test files
Source dataRecords textual triples expressing plausible relations among everyday events, social interactions, intentions, effects, wants, and reactions.
TermsCC BY 4.0
Unicode CLDR*
Public-distribution candidateSelected material48.2 common data
Source dataContains locale conventions for languages and territories, including names, calendars, date formats, number formats, and plural rules.
TermsUnicode License v3
Unicode Character Database*
Public-distribution candidateSelected material17.0.0
Source dataRecords Unicode character names, categories, scripts, case mappings, normalization properties, and text segmentation properties.
TermsUnicode License v3
IANA Time Zone Database*
Public-distribution candidateSelected materialtzdata 2026c, zone and rule data only
Source dataRecords time-zone identifiers, UTC offsets, daylight-saving transitions, and historical rule changes for named regions.
Selected materialv3.4.0
Source dataLists quantity kinds, measurement units, dimensions, prefixes, symbols, and conversion factors in formal vocabularies.
TermsCC BY 4.0
Selected materialRolling allCountries and hierarchy dumps
Source dataRecords global place names, alternate names, coordinates, feature classifications, country codes, and administrative hierarchies.
TermsCC BY 4.0
USGS Geographic Names Information System*
Tracked candidate, not yet pinnedSelected materialOptional national names file, exact revision pending
Source dataRecords official United States geographic feature names, locations, feature types, jurisdictions, and known historical or variant names.
NGA GEOnet Names Server*
Tracked candidate, not yet pinnedSelected materialOptional global names file, exact revision pending
Source dataRecords geographic names outside the United States, including approved and variant names, coordinates, feature types, and administrative codes.
TermsNo licensing requirements or restrictions; attribution recommended
Metamath set.mm*
Tracked candidateSelected materialdevelop at commit 18dfc1b2bd1a5db161aced29066199c783ec2926
Source dataContains formal axioms, definitions, theorem statements, and machine-checkable proofs in classical logic and ZFC set theory.
TermsCC0 1.0
Lean mathlib4*
Tracked candidateSelected materialmaster at commit 80ffd59621338133f22d4d6403537540e78aa68b
Source dataContains Lean definitions, mathematical structures, theorem statements, machine-checked proofs, tactics, and supporting programming infrastructure.
TermsApache License 2.0
License terms and review notes 17 sources
Asterisks on the cards jump to the matching row. Each row preserves the selected upstream terms, release caveat, and any source-specific license detail.
Swipe to review all license columns.
| Source and selection | License or terms | Review note |
|---|---|---|
| Wikidata entities and Lexemes Rolling entity and Lexeme dumps ↑ Card | CC0 1.0 | The exact downloaded snapshots and digests will be recorded in each release manifest. |
| Open English WordNet 2024 edition ↑ Card | CC BY 4.0 | The upstream notice also preserves the required Princeton WordNet attribution. |
| Open Multilingual Wordnet 1.4, permissive allowlist only ↑ Card | Per-lexicon licenses; selected members listed belowPer-lexicon license groups
|
The raw bundle is not uniformly CC BY. Only individually reviewed, permissive members may enter the friendly candidate. |
| Swedish FrameNet SweFN 2.0 ↑ Card | CC BY 4.0 | This is a separate source from Berkeley FrameNet and has its own terms. |
| PropBank Frames v2.1.5 ↑ Card | CC0 1.0 | The selected historical release is CC0. Later repository releases use ShareAlike terms, so this pin cannot move without a new review. |
| VerbNet 3.4 from the cu-clear repository ↑ Card | Historical VerbNet 3.X permissive grant; coverage under review | The official 3.X grant permits commercial use with notice preservation, but the selected repository does not carry its own license file. |
| Universal Dependencies Spanish AnCora r2.16 ↑ Card | CC BY 4.0 | UD licenses vary by treebank. The public pack's CC BY 4.0 terms apply only to the named treebank and release. |
| ATOMIC 2020 February 2021 train, development, and test files ↑ Card | CC BY 4.0 | The three selected files are one upstream dataset and will retain one source record plus file-level digests. |
| Unicode CLDR 48.2 common data ↑ Card | Unicode License v3 | The release manifest will pin the selected archive and preserve the Unicode notice. |
| Unicode Character Database 17.0.0 ↑ Card | Unicode License v3 | This is reviewed separately from CLDR even though both use the Unicode license. |
| IANA Time Zone Database tzdata 2026c, zone and rule data only ↑ Card | Public domain for selected data files | Separately licensed source-code files are outside the selected data scope. |
| QUDT v3.4.0 ↑ Card | CC BY 4.0 | The license applies to the selected QUDT repository release. |
| GeoNames Rolling allCountries and hierarchy dumps ↑ Card | CC BY 4.0 | Each fetch is a rolling snapshot, so its date and digest must be captured before release. |
| USGS Geographic Names Information System Optional national names file, exact revision pending ↑ Card | United States government work; public-domain posture | A dated source file and final scope review are required before this can become a release candidate. |
| NGA GEOnet Names Server Optional global names file, exact revision pending ↑ Card | No licensing requirements or restrictions; attribution recommended | The exact export and its governing notices must be captured before release review. |
| Metamath set.mm develop at commit 18dfc1b2bd1a5db161aced29066199c783ec2926 ↑ Card | CC0 1.0 | The selected file is scoped to public format and conformance work, not a published reasoning benchmark. |
| Lean mathlib4 master at commit 80ffd59621338133f22d4d6403537540e78aa68b ↑ Card | Apache License 2.0 | The selected revision is scoped to public format and conformance work. |
Boundary 03
Proprietary overlay
Public planning ledger revision 2026-08-23. A released artifact's signed manifest is authoritative.
What a release manifest will settle#
Each public pack is expected to carry a machine-readable manifest, a small file of facts about the pack itself, with:
- pack identity, version, and artifact digest (the checksum that fingerprints the download);
- exact upstream source and selected snapshot;
- applicable license or terms, attribution, and redistribution conditions;
- the public importer version used to produce the artifact, when one exists;
- enough provenance to trace public records to the selected source;
- compatibility and format-check status; and
- known exclusions or limits.
The manifest shipped with a pack will take precedence over summaries on this site. Organizations remain responsible for reviewing the manifest and their own obligations.
Downloads#
Public pack downloads are coming. Every source will have its own downloadable pack file; sources with different terms will remain separate. Public artifacts will carry their attribution, notices, version, digest, manifest, and format-check result.
The CleverMemory layer itself, the code and data that tie everything together, will not be offered as a pack download. That piece is not public; everything you need to check what it tells you is.
There are no public pack downloads today. Material changes will appear in Updates, and release notes will appear in the blog and the Atom feed.