Ternary Qualia+ Relations (TQR)
What is TQR?#
TQR (Ternary Qualia+ Relations) is FNBr's inventory of relations for characterising lexical meaning. Originally based on Pustejovsky's qualia structure, it is now extended with relations the four qualia don't cover, and grounds the whole set in the DUL foundational ontology.
The motivation is that representing context from linguistic material means associating what is expressed with commonsense knowledge — typically held in an ontology. Each LU carries a parcel of meaning that distinguishes it from other LUs evoking the same frame, and FNBr is building a lexical ontology to make that reasoning explicit. A lexical ontology explores what lexica and ontologies share while reducing the gaps between them. To implement it, FNBr reframed two ideas from SIMPLE: coding words as Lexical Units, and using "extended qualia relations" — now represented by Microframes. The full set of these relations is TQR.
Why Qualia+ ?#
Pustejovsky (1991, 1995) decomposes word meaning into four qualia — aspects that together characterise an entity:
| Quale | Question answered | Example (for pizza) |
|---|---|---|
| Formal | What kind of thing is it? | subtype_of Food |
| Constitutive | What is it made of? | is_made_of dough, cheese, tomato |
| Telic | What is it for? | object_of_the_activity eat |
| Agentive | How did it come to exist? | has_as_agent bake |
But the qualia answer questions about a single object's intrinsic profile — a noun-centric model. Even the SIMPLE inventory contains relations the qualia never accounted for:
- Ontological scaffolding — relations expressing ontological structure.
- Relational meaning between distinct entities — kinship, social roles, regulation. The meaning lives in the relation, not in either relatum's qualia.
- Eventive/temporal connective tissue —
precedes,follows,has_precondition,co-participates_with: how events sit in sequence and who participates, not what an object is for.
The "+" denotes exactly these non-qualia relations — spatial, eventive, social, and structural — that TQR brings under one inventory alongside the classic qualia.
DUL Integration#
To hold the extended set together, TQR integrates with DUL (DOLCE+DnS Ultralite). Microframes live in two namespaces: dul: for the foundational relations, and fnbr: for the lexical specialisations that subsume them. Each fnbr: qualia microframe carries at least one subsumption edge to a DUL microframe, grounding the lexical-semantic relation in a foundational ontology. This enables:
- Inferential generalisation — enables parser rules to fire on DUL relation families and propagate to all
fnbr:specialisations via the subsumption hierarchy. - Cross-domain interoperability —
fnbr:microframes are interpretable by any system that understands DUL object properties. - Unified modelling — frame-level and lexical-level relations share the same microframe abstraction.
The integration is non-breaking (adding subsumption edges does not alter existing frame structure) and descriptive, not prescriptive — the DUL alignment does not impose DUL's stronger ontological commitments on FNBr annotations.
Classification and grounding#
Every microframe — whether it grounds LU meaning through TQR or states a structural relation within a frame — carries a single user-facing semantic category plus a coarse grounding kind (property-grounded — a sub-property of a DUL object property — or description-grounded — a specialization of a DUL class such as dul:Plan), read off its DUL parent. Where applicable it also keeps its GL quale (Formal, Constitutive, Telic, Agentive) as provenance; a relation may have no quale (much of the scaffolding and eventive material does) while still having a category and a DUL parent.
This classification is a property of the microframe itself, not of TQR. It is described in full in the Microframes document (§ Classification), and how each grounding kind wires into DUL is covered in DUL Grounding.
TQR Structure#
The TQR data structure extends the previous version. Its core idea is that the relation between two LUs is grounded not only ontologically
but also in the FrameNet network. Since each microframe has a Domain and a Range, a TQR relation aligns each to a Frame Element of the
base frame it grounds; both FEs belong to that same base frame. This structure is called TQR_FE. It is distinct from rel_structure,
which attaches a microframe directly to two FEs of a frame without this LU-grounding layer (see Microframes § Use Cases).
A second structure, TQR_LU, records the individual relations between LUs; each TQR_LU points to two LUs and a TQR_FE (represented by the idEntityRelation
of Microframe-Frame rel_tqr).
Semantic Categories#
Microframes carry a single user-facing semantic category (13 values, type instances typ_microframe_category, idType 42). Because the category classifies every microframe — not only those used by TQR_FE — the full table lives in the Microframes document (§ Classification), and its grounding is detailed in DUL Grounding.