GUM: The Generalized Upper Model (OWL, GUM-3.1)¶
Local reference identifier:
xwkont:ref:gum-owlSlug:gum-owlEditorial status:candidateCreated:2026-07-02Modified:2026-07-05
Descriptive Metadata¶
| Field | Value | Notes |
|---|---|---|
| Title | GUM (Generalized Upper Model), version 3.1, OWL rendering | A linguistically-motivated ontology mediating between natural-language form and domain knowledge. |
| Creator | John A. Bateman | University of Bremen. |
| Contributor | unknown | Full contributor list not verified in this pass. |
| Publisher | University of Bremen, Faculty 10 (Linguistics/Language Processing group) | Originally self-hosted at fb10.uni-bremen.de; that host no longer resolves the ontology page (confirmed 404/dead redirect chain 2026-07-02 — see Source Relation Notes) and the canonical purl.org/net/gum2 PURL now redirects to a dead link. Retrieved via the Internet Archive Wayback Machine instead. |
| Source type | ontology specification (OWL) | |
| Version or edition | GUM 3.1 (GUM-31.owl, part of the GUM-3.owl namespace) |
An earlier GUM-3-space.owl module and a CASL rendering (GUM.casl) also exist at the same archived location but were not fetched in this pass. |
| Identifier or URL | Canonical PURL: https://purl.org/net/gum2 (currently redirects to a dead host chain, confirmed 2026-07-02); archived source: http://www.fb10.uni-bremen.de/anglistik/langpro/webspace/jb/gum/gum-files/GUM-31.owl.txt |
No DOI for the OWL file itself. |
| Access date | 2026-07-02 | |
| Snapshot / Stable identifier | http://web.archive.org/web/2021/http://www.fb10.uni-bremen.de/anglistik/langpro/webspace/jb/gum/gum-files/GUM-31.owl.txt (Wayback Machine, resolved to a 2021-dated capture at fetch time), verified retrievable at time of writing, per ADR-0012. The file was fetched through this archival URL directly (the canonical host no longer serves it live), so the snapshot and the fetched artifact are the same access, not a separate archival step. |
|
| Rights/license | unknown | Not stated in the fetched OWL file's header; not guessed. |
Source Relation Notes¶
GUM's canonical hosting has decayed. The purl.org/net/gum2 PURL (cited in ADR-0015 and TODO.md as GUM's "stable OWL DL identifier") now 307-redirects to purl.archive.org/net/gum2, which 302-redirects to the original fb10.uni-bremen.de page, which 404s (confirmed 2026-07-02) — the live chain is fully dead, not merely slow or paywalled. The Internet Archive Wayback Machine has a working capture of the original index page and its linked GUM-31.owl.txt file, which is what this record and continuant-occurrent.md cite. If GUM is later republished at a new stable location (e.g., a GitHub mirror), prefer that as the canonical source and retain this record as the historical/archival citation.
The dedicated 2022 Applied Ontology paper was not accessed. Bateman, J. A. (2022), "GUM: The generalized upper model," 10.3233/AO-210258, is hosted on the same platform that blocked xwkont:ref:borgo-galton-kutz-2022, the YAMATO 2022 paper, and the TUpper 2022 paper — not independently re-checked in this pass, but assumed blocked by the same pattern given three-for-three prior results.
Rights and License Notes¶
Not stated in the fetched file; do not assert a license. The original page (per its archived index) references academic/research use context typical of a university-hosted linguistics ontology, but this is not a substitute for an explicit license statement.
Citation and Locator Notes¶
Recommended citation: Bateman, J. A. GUM: The Generalized Upper Model, version 3.1 (OWL). University of Bremen. Archived at http://web.archive.org/web/20211231000000*/fb10.uni-bremen.de/anglistik/langpro/webspace/jb/gum/.
Direct primary-source verification (2026-07-02): GUM-31.owl was fetched directly (via the Wayback Machine, per the dead-canonical-host note above) and read directly to verify continuant-occurrent.md's GUM correspondence. GUM does not use continuant/occurrent, endurant/perdurant, or any comparable philosophical top-level label — its top-level split is a linguistic one, SimpleThing (nominal-group-realized participants) vs. Process (verb-realized "goings-on"), both subclasses of Element, explicitly owl:disjointWith each other. Exact rdfs:comment text:
SimpleThing(http://www.ontospace.uni-bremen.de/ontology/stable/GUM-3.owl#SimpleThing): "An entity which may participate in a configuration." Subclass ofElement.Process(http://www.ontospace.uni-bremen.de/ontology/stable/GUM-3.owl#Process): "A Process is the linguistic construal of 'goings-on' or events. Processes are similar to Configurations, but factor out Participants, Circumstances, and other elements. As such, entities classified under Process can be expressed as verbs and are frequently the main verb in a clause." Subclass ofElement;owl:disjointWithSimpleThingandSimpleQuality.GUMThing(http://www.ontospace.uni-bremen.de/ontology/stable/GUM-3.owl#GUMThing), the ontology's actual top node: "The most general experiential category that can be expressed through the resources of the lexicogrammar of a language. UMThing corresponds to 'phenomena' in Halliday and Mathiessen (1999: 48)." [sic— "UMThing" in the source text, likely a typo for "GUMThing" carried over from an earlier "UM" (Upper Model) naming.]
GUM's SimpleThing/Process split is grounded in systemic-functional linguistics (Halliday & Matthiessen), not in a metaphysical persistence criterion — the closest structural analog to BFO's continuant/occurrent in this crosswalk's existing SUMO precedent (note-003), not a terminological match. See continuant-occurrent.md's Uncertainty section.
Reused for object.md (2026-07-02): DecomposableObject ("An object that is being viewed as being made up of parts that may be taken apart and are often given explicit linguistic recognition.") and NonDecomposableObject ("An object that is being regarded as not possessing significant parts, or which is not to be considered decomposable for present purposes."), both direct subclasses of SimpleThing and mutually owl:disjointWith, are GUM's closest analog to a single "Object" category. See object.md's correspondence rows 009-010.
Reused for event.md (2026-07-02): GUM's Process class definition (quoted above: "the linguistic construal of 'goings-on' or events") is the only one of the four ADR-0015 sources besides BFO's skos:altLabel usage to use the literal word "events" directly in a class definition — but, like BFO, as an informal gloss on a general occurrent-side class, not a separate named Event category. See event.md's Source Definitions and correspondence row 010.
Reused for process.md (2026-07-02): the same Process class (quoted above) is also the only one of the four ADR-0015 sources with a class literally named "Process" — the strongest terminological match to this crosswalk's subject of any new source, though its criterion (verb-realizability) is linguistic rather than temporal/metaphysical. See process.md's Source Definitions and correspondence row 008.
Reused for information-artifact.md (2026-07-02): GUM-31.owl was searched for an Information-Object-equivalent entity class; none was found. The only relevant hooks, Symbolization (no rdfs:comment) and its subclass Signification ("linguistically realized by verbs like 'represent', 'mean', 'express', e.g., 'green means go'"), are relation types (subclasses of Intensive), not object/entity types — considered and rejected as a correspondence. See information-artifact.md's Source Definitions.
Reused for quality.md (2026-07-02): GUM-31.owl also supplies SimpleQuality ("Qualities are properties of SimpleThings and Processes... anything that can be expressed as an English adjective"), subclass of Element, owl:disjointWith SimpleThing. The only one of the four new sources with a class literally named "Quality" (via subclasses like MaterialWorldQuality/ModalQuality/LogicalQuality), though grounded in adjectival/linguistic realization rather than metaphysical inherence. See quality.md's Source Definitions and correspondence row 007.
Reused for relation.md (2026-07-02): GUM-31.owl also supplies Relating ("A configuration indicating a relation holding between two entities. A relating must contain at least a domain and an attribute"), rdfs:subClassOf BeingAndHaving, itself rdfs:subClassOf Configuration (a clause-level situation construct containing a Process, Participants, and Circumstances). A fourth distinct relation-modeling style — reified, but as a situation/state-of-affairs, not a trope, abstracta, or bare predicate. See relation.md's Source Definitions and correspondence row 007.
Reused for role.md (2026-07-02): GUM-31.owl also supplies RolePlaying ("A circumstantial relationship that expresses a restriction in which a facet of one of the participants in a process is relevant for the actualization of the process... frequently realized in English by a prepositional phrase with the preposition 'as'; For example: 'As a president, he was terrible, although as a golfer he was not too bad.'"), rdfs:subClassOf Symbolization. A relation type, not an object/entity type — the third concept in this repository (after Information Artifact's Signification/Symbolization and Relation's Relating) where GUM reifies this kind of content as a relation rather than an entity. See role.md's Source Definitions and correspondence row 007.
Reused for time.md (2026-07-04): GUM-31.owl was fetched fresh (via curl through the same Wayback Machine snapshot — WebFetch itself cannot reach web.archive.org in this environment, confirmed again this pass) and found to have by far the richest temporal treatment of any of the 8 sources: 18 distinct classes under a TemporalObject/Time/TimeInterval/TimePoint hierarchy (full tree in time.md's Source Definitions), including Past/Present/Future, TemporallyBounded/TemporallyUnbounded, DurationInTime/InstantaneousInTime, OneOrTwoDTime/ThreeDTime, and ProfilesAndStages/TemporalProfile. Unlike every other source checked, GUM's temporal apparatus is explicitly grounded in Halliday & Matthiessen's (1999) systemic-functional-linguistics theory of the clause, cited by page number directly inside multiple class definitions (e.g. DurationInTime: "A Configuration may be take time to unfold or be instantaneous (HM 1999: 471)") — modeling how a clause's temporal profile is grammatically realized (tense, aspect, boundedness), not time as a mind-independent feature of reality. Time itself has a dual parent, rdfs:subClassOf both Substance and TemporalObject — neither of GUM's occurrent-like (Process) or continuant-like (SimpleThing/DecomposableObject) classes are ancestors of any temporal class. This corrects docs/evaluations/foundational-ontology-concept-terms-matrix.md's Observation 2, which omitted GUM from its list of sources with an explicit Time category entirely. See time.md's Source Definitions and correspondence row 008.
Reused for docs/methodology/source-ontology-module-conventions.md (2026-07-05, checking ADR-0021's GUM gap): this record's own Descriptive Metadata already notes an unfetched GUM-3-space.owl module existing alongside GUM-31.owl at the same archived location — suggestive of some modular structure, but a secondary-literature search of Bateman's own GUM papers found no "core" terminology or explicit base/extension description for GUM's own architecture. Recorded as inconclusive, not a confirmed absence — GUM's canonical host is dead (see Source Relation Notes above) and a full site fetch was not attempted in this pass.
Reused for list-sequence.md (2026-07-07): GUM-31.owl was re-fetched and found to supply OrderedObject ("A type of decomposable object whose parts have an intrinsic ordering of their own; for example, the elements of a list, the carriages of a train, etc. (Used to be \"Element\".)"), rdfs:subClassOf DecomposableObject, and its narrower subtype OrderedSet ("A set whose elements are ordered."), rdfs:subClassOf both OrderedObject and UMSet. This is a genuine second source for that crosswalk's ordered-collection candidate, missed entirely by an earlier drafting pass that checked only SUMO's Merge.kif. Also confirmed a separate ElementList class (rdfs:subClassOf Part), a part-whole relation type between a list and its elements, not an entity class — filed under mereology-parthood-aggregate.md's territory instead, not list-sequence.md's. See list-sequence.md's Source Definitions and correspondence row 002.
Reused for continuous-discrete.md (2026-07-07): GUM-31.owl was re-fetched and grepped directly for "continuous," "discrete," "granular," and "atomic." Zero hits for any of the four terms in the file. Confirms the crosswalk's draft finding of a clean absence exactly. See continuous-discrete.md's Source Definitions.
Reused for modality.md (2026-07-07, this candidate's source-side anchor): GUM-31.owl was re-fetched and all eight Modality-tagged classes (candidate-concepts.md's GUM.28/98/99/107/113/118/133/188) were read directly. Confirmed a coherent two-tier tree: ModalQuality (rdfs:subClassOf SimpleQuality, "Qualities of being able to do something, wanting to do something, having to do something, and so forth") is the root; Conditional, NonConditional, NonVolitional, and Volitional are its direct children (mutually owl:disjointWith in various combinations); Necessity and Possibility sit one level lower, rdfs:subClassOf NonVolitional (owl:disjointWith each other). A separate but directly related class, ModalPropertyAscription (rdfs:subClassOf PropertyAscription, restriction: attribute owl:allValuesFrom ModalQuality), is the ascription relation pairing a ModalQuality to a process — structurally parallel to ColorPropertyAscription/SizePropertyAscription/etc. Also confirmed during review: PropertyAscription (the shared superclass) itself is owl:disjointWith only QuantityAscription, not any sibling *PropertyAscription class; the additional disjointness between ModalPropertyAscription and LogicalPropertyAscription is declared reciprocally on LogicalPropertyAscription's own class block, not inherited from the superclass or declared on ModalPropertyAscription's own block. Two further Possibility subclasses, GeneralPossibility and Ability, are structurally inside this same tree but out of this crosswalk's scope — already tagged Disposition / Capacity in candidate-concepts.md and treated in disposition-capacity.md. See modality.md's Source Definitions, correspondence row 008, and uncertainty-005.
Reused for non-physical-social-object.md (2026-07-07, session-048): GUM-31.owl was re-fetched and SimpleThing's class block read directly — confirmed the draft's quotation exactly: owl:Class#SimpleThing, rdfs:subClassOf Element, rdfs:comment "An entity which may participate in a configuration." No non-physical/social-object-specific hierarchy exists beneath it; GUM's split is the linguistic SimpleThing/Process one already documented for continuant-occurrent.md. See non-physical-social-object.md's Source Definitions and correspondence row 008.
Reused for symbol-sign-representation.md (2026-07-07, session-049), independent re-verification found and corrected two truncated quotations: GUM-31.owl was re-fetched fresh (via the same Wayback Machine snapshot) and the Name, NameEvent, NameOf, NameRelation, Symbolization, and Signification class blocks re-read directly. Name ("An accepted but possibly arbitrary label for some entity.") and NameOf ("The relation that holds betwen a name and its bearer." — [sic], the typo is in the source file itself) matched the prior pass exactly. Two quotations were found truncated in the 2026-07-06 draft and corrected: Symbolization's full rdfs:comment reads "This category contains relations that hold between entities and other entities they symbolize'. Concepts in this class logically have two roles: SYMBOL (the domain) and SYMBOLIZED (the range). One relation in this category is the relation between an object and a name which identifies that object." — the prior pass quoted only the first clause, dropping the domain/range framing and worked example.Signification's full comment reads "Relations concerned with Signification are linguistically realized by verbs like 'represent', 'mean', 'express', e.g., 'green means go'." — the prior pass dropped the "linguistically realized" framing and the "green means go" example.NameEvent's comment was also found to carry a second sentence not previously quoted: "The process of giving something a name. It does not intrinsically require an addressee (non-addressee-oriented)." All four full quotations now recorded insymbol-sign-representation.md`'s Source Definitions.
Reused for information-artifact.md and event.md (2026-07-07), checking symbol-sign-representation.yaml's own Future Work items: GUM-31.owl was re-fetched fresh and the Name and NameEvent class blocks re-read directly. Name ("An accepted but possibly arbitrary label for some entity.") is rdfs:subClassOf Word, itself rdfs:subClassOf Abstraction — a genuine object/entity-level class, unlike information-artifact.md's already-rejected Signification/Symbolization relation types; added there as correspondence-005/mapping-009. NameEvent's ancestry was traced beyond the previously-quoted DispositiveMaterialAction/External pair: DispositiveMaterialAction → AffectingAction → DoingAndHappening → Configuration, a structurally distinct branch from event.md's existing GUM correspondence (Process) — Configuration requires exactly one Process component plus factors in Participants/Circumstances that Process alone factors out. Added to event.md as correspondence-011/mapping-011, with the Process-vs-Configuration weighting question left open as uncertainty-007.