The Ontological Dimension
FrameNet Brasil records which frame a lexical unit evokes and how its frame elements are realized in sentences. The ontological dimension makes the resource's semantic categories explicit: what a lemma names, what a frame describes, and what kind of concept or participant a frame element represents.
This classification supports review and reasoning; it does not replace lexical meanings, frame definitions or human annotations. An automated compatibility flag is a request for review, not a semantic verdict.
Lexical and frame identity#
- A lemma is a lexical entry, distinguished by language, citation form, POS and, where needed, homograph identity. It carries a conceptual or procedural type.
- A lexical unit (LU) represents a reading of a lemma associated with a frame, with a sense description and lexical contribution. Several LU records may require review for duplication or mixed readings; a shared lemma/frame pair alone does not prove equivalence.
- A frame describes a situation or kind of referent and its participant structure. Each accepted frame has exactly one namespace.
- An annotation set records an interpretation of a target occurrence and its realized or implicit FEs. A stored annotation is evidence, not automatic approval of the interpretation.
A lemma can have multiple readings, and those readings can evoke frames in different namespaces. This is normal polysemy, but each allocation must still fit the particular reading. Namespace diversity neither proves an error nor excuses one.
The type is recorded on the lemma, not independently on each LU. Type, namespace and grammatical aspect answer different questions; see Lemmas and Frames.
Event, Situation and Description#
DUL's Descriptions & Situations pattern distinguishes:
| DUL notion | Meaning | Use in interpreting FrameNet |
|---|---|---|
| Description | An intensional account defining concepts and roles | A frame supplies a description against which an occurrence or referent is interpreted. |
| Concept | A social object defined in some Description | A Frame Element defined in a Frame. |
| Situation | A contextualized setting satisfying a description | An annotation records an analysis of a described situation. |
| Event | Something that occurs | An event-denoting target can refer to an event viewed through a frame. |
These are semantic correspondences, not identities between database records and things in the world. An LU is not a DUL Event, and an annotation set is not the event itself. A lexical reading may be used for many different events; Entity, Attribute and Relation readings need not denote events at all. Fictional, hypothetical and negated sentences do not establish actual occurrence.
One happening may support several descriptions: João quebrou o vaso profiles causation, while o vaso quebrou profiles the transition. Their LU meanings and frame distinctions remain separate. Passive voice alone does not create a different frame or namespace.
See DUL design patterns for the grounding vocabulary and FrameNet patterns for lexical/frame separation criteria.
Independent classification axes#
| Object | Classification question | Documentation |
|---|---|---|
| Lemma | What conceptual kind or processing instruction does this lexeme encode? | Lemma types |
| Frame | What situation, referent or modeling function does the frame describe? | Frame namespaces |
| FE | What concept-kind and semantic role does this FE have? | Frame elements |
| FE filler | What may occupy or be interpreted in the FE slot? | Filler restrictions |
| Microframe | What reusable binary relation is represented? | Microframes |
A role's own kind differs from the class of its fillers: Cook is a role, while its filler may be a sentient person. A frame's namespace does not follow from the filler type of one FE.
DUL grounding supplies a shared hierarchy. Use informative classes rather than the uninformative root Entity where a more specific class is justified. FNBr namespace and lemma-type inventories are local modeling devices, not inventories prescribed by Berkeley FrameNet or DUL.
Namespace, perspective and grounding#
These devices are related but not interchangeable:
- Namespace classifies a frame as a whole. Some eventive categories encode coarse participant perspective; other categories distinguish entities, properties, relations or scaffolding.
- Perspective_on relates a perspectivized frame to a neutral description of the same scene. It is not a relation between every pair of frames sharing a namespace.
- LU meaning specifies the lexical reading and contribution. A frame cannot impose an arbitrary sense on a word merely because its type is compatible.
- Conceptual grounding relates a frame to a schema. The schema supplies structure, not proof of correct namespace or LU allocation.
Preserve causative/inchoative/state, neutral/perspective and event/condition distinctions. When definitions, FEs and examples disagree, record the conflict and defer unsupported corrections instead of inferring a verdict from names, morphology or majority type counts.