Grounding a Frame in a Schema

Baseline: prod, 2026-10-13, read-only. No explicit rel_grounded_in row touches the 45 Structural frames or their 168 local FEs in the inspected snapshot. This page therefore distinguishes a modeling method from proof of populated production groundings. See the evidence and scope.

What a grounding analysis should establish#

To analyze a lexical frame against a schema, identify:

The lexical frame's defined situation, essential participants and relations.
The schema's actual local FEs and structure, using stable frame/FE IDs.
Which lexical roles correspond to which schema roles, and why their meanings match.
Which aspects remain unrepresented or need another structure.
Whether the analysis is a proposal or has explicit stored grounding rows and FE mappings.

A matching name is not enough. A lexical Ground can be a reference object, while CHANGE.Ground names a dimension. Conversely, different FE names may express compatible contributions. Neither namespace membership nor the number of roles selects a schema automatically.

The previous instruction to choose “the most specific schema” is a useful preference only when that specialization actually fits. Do not invent restrictions or bindings to reach a preferred leaf.

Read the relation type before interpreting a link#

Relation	Stored direction / structure	Interpretation
`rel_inheritance` (1)	Parent frame → child frame; attached FE rows use the same direction.	Specialization, not a Grounded_in edge.
`rel_using` (3)	Used/base frame → using frame.	Reuse of background structure; not automatically grounding or inheritance.
`rel_sub_frame` (308)	Composite frame → component frame; attached local FE → component FE mappings.	Schema composition in the current inventory. Distinct from `rel_subframe` (2).
`rel_grounded_in` (245)	Grounding source → grounding target.	Explicit grounding; absent from the captured Structural neighborhood.
Frame constraint (`cng_frame`)	A constrained FE → constraining frame.	Filler/coercion evidence for that FE, not a relation between the host frame and a schema.
`rel_structure` (265)	Two local FEs plus a mediator frame; attached role bindings.	A local assertion; see Microframes.

For example, MOVEMENT_ALONG_A_PATH.Ground has an explicit frame constraint to PATH. That is not a stored Grounded_in edge from MOVEMENT_ALONG_A_PATH to PATH. The distinction matters when checking coverage or counting groundings.

A verified composition example#

LOCOMOTION (4356) has one local FE, Mover. The database stores:

Composite → component	Frame relation ID	Attached FE mapping
LOCOMOTION → MOTOR_CONTROL (4351)	5416326	Mover → Agonist
LOCOMOTION → MOVEMENT_ALONG_A_PATH (4335)	5416328	Mover → Protagonist

The shared Mover connects the controlled actor and moving participant. This is observable composition evidence, not an inferred extra FE and not two Grounded_in edges. Its separate inheritance parent is EVENT, but no attached EVENT → LOCOMOTION FE inheritance mapping is present in this snapshot.

FORCEFUL_MOTION_ACTION provides a larger verified example: three direct components (FORCE_APPLICATION, EFFECTOR_SPATIAL_ACTION, MOVEMENT_ALONG_A_PATH) and seven mappings. The inventory lists every composition edge.

Combining analyses without conflating them#

Motion and containment can both be relevant to an entering frame. Such a proposal must explain what each structure contributes and what evidence would support each mapping. It is not certified merely by the word entrar or by the availability of MOTION and CONTAINER.

When reviewing multiple structures, inspect coverage across the proposed combination, but do not assume that all core roles are currently covered. The snapshot contains 21 internal inheritance edges without attached FE mappings, and no explicit Grounded_in rows for the selected neighborhood. A frame relation alone does not demonstrate a complete role mapping.

Profiles are not automatically validated bindings#

The former page said that a frame's profile was automatically read off core FE bindings and that an empty profile intentionally looked identical to an unanalyzed frame. Those statements should not be treated as evidence of a working production pipeline.

For review, distinguish at least:

Not analyzed: no supported mapping recorded.
Partially analyzed: some roles mapped, others unresolved.
Analyzed with a specified profile: explicit semantic decisions with traceable evidence.

A local FE's coreness, an LU's lexical rank and a schema role's participation are different facts. Core-Unexpressed is not simply “optional.” Peripheral mappings do not automatically establish an LU profile. See Lexical Units and Profiling for the repository's Element-centered lexical profile contract.

All local schema FEs in this production snapshot have idElement = NULL; the documentation refresh neither creates Elements nor migrates mappings. Absence of such links here is not a full audit of lexical profiling elsewhere.

Review record#

Before proposing a new grounding, record the frame and schema IDs, intended relation type and direction, FE correspondences, unmatched roles, definition/example evidence and competing analyses. Keep proposed, stored and validated status separate. The diagnostics help organize that review; they do not authorize automatic data changes.