Type classification for lemmas

Every lemma carries exactly one type. The type vocabulary has two sorts: ontological types, for lemmas that encode a concept, and procedural types, for lemmas that encode an instruction for processing. Which sort a lemma takes is the first decision the annotator makes; everything else follows from it.

A lemma here is a lexeme: a citation form paired with a part of speech, not a bare string. A lexical unit (LU) is a lemma paired with a frame, carrying a sense description — the relation is N:M, exactly as in Berkeley FrameNet. The type belongs to the lemma and only to the lemma. The LU inherits it and carries no type of its own, so creating an LU involves no typing decision at all.

What this layer claims, and what it does not#

Four scoping commitments govern every assignment in this document. They are stated first because most objections to a typing layer of this kind are objections to claims we are not making.

1. The type is a property of the lemma — a lexeme, text plus POS — not of a word form and not of the referent. FrameNet's own position is that the lemma is already a meaningful unit: it is why lemmas carry POS at all, and why the LU table needs to record nothing beyond the lemma–frame pairing and a sense description. A lexeme is not construal-indeterminate the way a bare string is: colar/VERB and colar/NOUN are different lexemes with different meanings, and the POS has already done that separating work before any frame is consulted.

2. The type classifies a conceptual profile, not the furniture of the world. No claim is made about what exists. This is consistent with our grounding ontology: DOLCE is a Descriptive Ontology for Linguistic and Cognitive Engineering, an inventory of particulars as conceptualized by natural language and common sense, explicitly neutral on metaphysics. Typing a lemma against DUL is typing a conceptualization against a catalogue of conceptualizations.

3. The type is a coarse, defeasible default, serving as the input to constructional composition and coercion. A construction may override it. Override is the interesting data, not a counterexample: coercion is a function, and a function needs an argument — if the lexical item brought no type, there would be nothing for the construction to override. Locating the default on the lemma is what makes this commitment coherent: the default lives in the lexicon and is defeated in use.

4. The type is a comparative concept, in Haspelmath's sense: an FNBr-internal instrument for cross-frame, cross-resource and cross-linguistic alignment. It is not a descriptive category of Portuguese, and it is not a metaphysical claim.

Commitment 3 has a cost, recorded here so it is not discovered later: because the type is defeasible, the class-compatibility check in §10 flags a marked combination worth review, not an error. We cannot have both defeasibility and hard validation, and defeasibility is the more defensible of the two.

1. Parts of speech and the locus of the type#

1.1 Why retain POS at all#

The theoretical case against parts of speech is strong and we accept it: syntactic categories are language-particular and construction-derived (Croft, Radical Construction Grammar 2001; Morphosyntax 2022), and cannot be equated across languages as descriptive categories (Haspelmath 2010). This is why the type layer in this document exists and why it is not a POS layer.

It does not follow that POS should be discarded. POS remains load-bearing for work FNBr actually does: lemmatization and morphological analysis for a morphologically rich language, MWE and terminology identification, corpus query (word sketches are defined over POS patterns), and interoperability with UD, LMF and OntoLex-Lemon/LexInfo. Discarding it breaks all of that and buys nothing, because the ontological type was never going to do that work.

The correct architecture is the one OntoLex-Lemon already prescribes: POS is morphosyntactic metadata on the lexical entry; the type is the conceptual layer, linked by semantics-by-reference. Both sit on the lemma; neither pretends to be the other.

1.2 POS is part of lemma identity#

POS is a property of the lexical entry = (lemma string + POS), and the lemma key must include it:

lemma(id, text, pos, lang, homograph_idx, type_id)
        UNIQUE(text, pos, lang, homograph_idx)
form (id, text, lemma_id, feats)      -- the paradigm of ONE lemma entry
lu   (id, lemma_id, frame_id, sense_description)

Keying lemma on text alone conflates distinct lexemes and leaks one paradigm into another. Colar is two lemmas: colar/VERB ("to glue"), whose paradigm contains colo, colei, colando; and colar/NOUN ("necklace"), whose paradigm is colar, colares. With a text-only key, a parse of Eu colo o papel resolves colo to a single colar row and the noun sense becomes reachable from a verb form — a lemma that does not exist.

Two consequences follow:

• Form lookup is inherently set-valued. The string colar is simultaneously the infinitive of colar/VERB and the singular of colar/NOUN. form.text → {lemma_id} returns candidates; the UPOS from the UD parse filters them. Colo belongs to only one paradigm, so the noun stops being offered — with no semantic reasoning involved at all.
• When UPOS is absent or unreliable, do not guess at the lemma layer. Return the candidate set and let frame matching decide. POS narrows; it does not adjudicate.

homograph_idx distinguishes lemmas that share text and POS. It covers two cases: genuine homonymy (manga fruit / sleeve), and type divergence, where one form under one POS carries readings of different types. Default 0. It is not used for polysemy within a single type, which is what LUs are for. See §1.4 for the rule.

For MWEs, pos is nullable and entry_type ∈ {word, mwe, affix} carries the distinction. UD has no honest UPOS for dar um jeito, and forcing the head's tag into that column corrupts the lookup index. This is also where the type layer visibly outperforms POS: an MWE has no coherent part of speech but has a perfectly coherent ontological type.

1.3 Why the type belongs to the lemma#

Three reasons, one theoretical and two practical.

The lexeme is already a form–meaning pairing. Once POS is part of lemma identity, most of the construal-fixing work is already done before any frame is consulted. Colarᵥ and colarₙ are separate lemmas with separate types; so are dependenteₙ (role) and dependenteₐ (value). Attributing the type to the lemma–frame pairing instead would record on every LU a fact that is constant across all of a lemma's LUs.

The type is then available before frame assignment. A parser or lemma-identification pipeline can type a token as soon as it has resolved the lemma, with no frame in hand. If the type lived on the LU it would be unreachable until frame disambiguation had already succeeded — which is precisely when it would have been most useful as a filter.

It matches the FrameNet data model. The LU table is a lemma × frame association with a sense description. Adding a mandatory type column to it would make the LU a second lexical entry rather than a pairing.

1.4 One lemma per (text, POS, type)#

Where one form under one POS carries readings of different types, each type gets its own lemma row, distinguished by homograph_idx. Readings that share a type stay in one lemma and are distinguished as LUs.

One lemma per (text, POS, type).

This is the only mechanism for type divergence in the specification. There is no LU-level type and no override. Two consequences follow, and both are the reason for the rule:

• Creating an LU involves no typing decision. The annotator pairs a lemma with a frame and writes a sense description. The type arrived with the lemma, was decided once, and is not re-litigated per frame.
• lemma.type_id is correct by construction. It is non-null, single-valued and never wrong for any LU beneath it, because a divergent reading is by definition a different lemma.

The second clause of the rule keeps the split bounded. De has at least six distinguishable readings but only two types, so it yields two lemmas, not six (§8.2). A lemma splits on type, never on sense count.

Three case shapes are known:

• Cross-sort polysemy. Negação names the act of denying (event, in an @agentive frame) and the propositional operator (polarity). One form, one POS, two sorts → negação₀ event, negação₁ polarity.
• Product nominalization. Pintura as the activity is event; pintura as the canvas hanging on a wall is object. The same holds for construção, tradução, invenção, criação, publicação — deverbal nominals whose event produces an artifact that outlives it. Not all deverbal nominals: see §5, "event / object in deverbal nominals".
• Adpositions. De, a, em, por, com, para, sobre. The largest group, worked through in §8.2.

Two things the rule is not for:

• Not for determination-driven shifts. Pai used to pick out an individual who walked in, with no relation in play, is still the role lemma under a referential shift. Löbner shows such shifts are regular and productive; per commitment 3 they are defeasance in use, not a second type. No split.
• Not for polysemy within a type. Genitive, material and partitive de are one predication lemma with several LUs. Two frames do not imply two lemmas; two types do.

Splitting is the same operation as homonymy handling, and deliberately so: manga fruit/sleeve and de₀/de₁ produce identical rows. The specification does not ask annotators to decide whether two readings are "the same word", because that judgement is unreliable and, once the type is settled, has no consequence for the data.

What a split does not record. Two lemmas carry no relation to each other. Where the relatedness matters — negação act/operator, pintura activity/canvas — it belongs at the lexical layer as an explicit link between lemmas, which is what OntoLex-Lemon's vartrans module is for. It should not be implied by a shared row (§12.7).

1.5 POS and type are not orthogonal, and we do not claim they are#

Earlier versions of this document said the type is "orthogonal to the UD POS tag." That overstates the relationship. Deverbal nominalizations and participles will show strong type/POS correlation, because category is assigned in a structure that also determines event structure. The defensible claim is weaker and sufficient: the type is not derivable from POS and not reducible to it. How much independent information it carries is an empirical question, answerable by a POS × type contingency table once both columns are populated — and now trivially computable, since both live on the same row.

2. The two sorts: conceptual and procedural encoding#

Not every word encodes a concept. Quem, não, portanto, o have meanings, but those meanings are not things that could be classified as objects, events or qualities — the "world thing" is computed in context, if at all. Forcing them into an ontological inventory is what turned .relation into a wastebasket in the previous version of this specification, sitting parentesco next to não and portanto.

Relevance Theory supplies the distinction we need. Blakemore (Semantic Constraints on Relevance, 1987) and Wilson & Sperber ("Linguistic form and relevance", Lingua 1993) distinguish conceptual encoding — the expression contributes a constituent to the proposition expressed — from procedural encoding — the expression encodes an instruction for processing: a constraint on inference, on reference resolution, on how to build or access a representation. Discourse connectives, indexicals, polarity, and most grammatical morphemes are procedural on this analysis.

FNBr adopts the distinction as the top branch of the type vocabulary:

Sort	Encodes	Gets an ontological type	Gets a DUL reference
conceptual	a concept	yes (§4)	yes
procedural	an instruction for processing	no	no

Procedural lemmas carry no DUL reference, and this is the honest position rather than a gap: there is nothing in the conceptualization inventory for them to reference, because they do not name a conceptualized particular. They name an operation on one.

3. Definitions#

The diagnostics in this document use six notions from formal ontology and lexical semantics. They are defined here once.

Rigidity (Guarino & Welty, OntoClean). A property is rigid if it is essential to all its instances: anything that is P must be P for as long as it exists (pessoa, cadeira). It is anti-rigid if every instance could cease to be P and survive (comprador, criança, cansado). It is semi-rigid if essential for some instances and accidental for others. Anti-rigid properties cannot subsume rigid ones — which is why pessoa is not a kind of agente.

Rigidity is defined here but is not a typing axis — see §4, "Why rigidity is not a typing axis". It remains useful as background for reading the taxonomy and for any future export to a rigidity-aware ontology.

Identity and sortality. A property supplies identity if it carries a criterion for saying "the same one again". Properties that supply identity are sortals; they answer what something is. Pessoa supplies identity; comprador inherits it from pessoa; vermelho has none. Every object instantiates exactly one identity-supplying kind.

Unity. A property carries unity if all its instances are wholes under one and the same unifying relation (functional, topological, morphological, intentional). Pessoa carries functional unity; água is anti-unity — an arbitrary sub-portion of water is still water. Unity is what separates object-like from substance-like concepts; FNBr does not currently type on it, but the notion is used in §4 to explain why mass nouns need no separate type.

External dependence. A property is externally dependent (+D) if every instance requires the existence of some entity y that is not a part or constituent of it. Comprador requires a purchase and a seller; filho requires a parent; cadeira requires nothing. The "not a part" clause is what makes the dependence external, distinguishing it from the trivial dependence of a whole on its parts. This is the diagnostic that replaces constitutivity (§5).

Relationality and uniqueness (Löbner, "Concept Types and Determination", Journal of Semantics 28(3), 2011). A noun concept is relational [+R] if it has an inherent argument slot for a relatum (irmã, filho), and unique [+U] if it determines at most one referent, absolutely or relative to its relatum (sol; pai relative to a child). Crossing the two yields sortal [−R−U], individual [−R+U], relational [+R−U] and functional [+R+U] concepts. FNBr does not type on [U], but both features supply corpus- observable diagnostics (§5) where the ontological test alone is hard to apply.

Dimension and region (Gärdenfors, Conceptual Spaces; The Geometry of Meaning). A quality is a dimension — an axis if gradable, a class if categorial. A region is a value the dimension can take. Temperatura is a dimension; quente is a region within it. This is the distinction underlying quality vs value.

4. The ontological types#

Six types. Each grounds in a DUL class. There are no parameters: the previous .entity<event> / .entity<relation> / .attribute<value> / .state<value> devices are retired in favour of named types (§11).

Type	DUL grounding	Criterion	Examples
event	dul:Event	exists in time; occurs or happens	destruição, aquecer, correr, quebrar, derreter, existir
object	dul:Object	a sortal with no external dependence	cadeira, água, pessoa, documento, lei, ideia, criança, adulto
role	dul:Role ⊑ dul:Concept	a sortal that is externally dependent	comprador, jogador, cliente, filho, pai, proprietário
relation	dul:Relation ⊑ dul:Description	a reified relation type; defines the roles that fill its argument positions	parentesco, paternidade, posse, casamento, adjacência, anterioridade
quality	dul:Quality	a dimension inhering in a bearer	cor, tamanho, beleza, inteligência, temperatura, cansaço, postura
value	dul:Region ⊑ dul:Abstract	a region or point within a quality's dimension	azul, quente, bonito, inteligente, cansado, quebrado, sentado, belamente

The reference principle#

The lemma references a conceptualization, never a particular, and never a formal object. The principle holds in three places and should be stated once:

• comprador references dul:Role — a Concept that classifies persons — not any person.
• parentesco references dul:Relation, a Description — the relation type — not any particular kinship instance holding between two people.
• adjacência and anterioridade reference the commonsense dul:Relation, not the formal counterpart under dul:FormalEntity.

This is what "semantics by reference" means for a lexicon, and it is why the philosophical worry about typing words as though they were things does not arise: the lemma points at a conceptualization, which is exactly what a word does.

role and relation are two halves of one structure#

dul:Relation is a Description; the Concepts filling its argument positions are dul:isDefinedIn that Description. So:

FNBr object	DUL	DnS relation
relational frame	dul:Relation ⊑ Description	—
FE of that frame	dul:Role / dul:Concept	dul:isDefinedIn → the Description
relation lemma (parentesco)	references the Relation	—
role lemma (filho, pai)	references a Role	isDefinedIn the same Relation

Parentesco names the description; filho and pai name Concepts it defines. They are not two unrelated primitives that happen to sound alike.

The instance layer, where it is needed, is dul:Situation: the particular fatherhood holding between two individuals in an annotated sentence is a Situation that dul:satisfies the Relation. Lemma layer → Description; annotation layer → Situation. This is the same coarse-identity / fine-situation asymmetry the whole model runs on (§9).

Converse pairs now fall out of the structure. The previous version noted that relation-base factoring would make pai / filho sibling projections of a single base, and treated that as a reason one might one day revisit the design. With role and relation both grounded in DnS, the benefit is obtained without the re-factoring: both roles are isDefinedIn the same Relation, and the converse pairing is readable off that link.

Why rigidity is not a typing axis#

An earlier version of this specification had a seventh type, phase, for anti-rigid but intrinsic sortals: criança, adulto, novato, defunto — concepts an individual can stop falling under while continuing to exist, without any second entity being required. It was distinguished from object by rigidity and from role by dependence. It has been folded into object, and rigidity is no longer used to type anything.

The reason is annotator reliability. The dependence test asks a single existence question: is there some entity, not a part of this one, that must exist? The rigidity test asks a modal one: could this individual cease to fall under this concept and still exist? Modal judgements of that kind are exactly what made the old state/attribute boundary unreliable (§6), and dropping rigidity here is the same decision applied consistently rather than a concession made once.

What is not lost. The merge is consistent with DUL, not a departure from it: dul:Concept ⊑ dul:SocialObject ⊑ dul:Object, so a phase-concept falls under dul:Object whether or not we name it separately. Using the general class where we previously used a more specific one is a loss of specificity, not an error. The DnS structure of §4 is untouched, since phase never participated in it.

It also repairs role. With two tests, role required anti-rigidity and dependence, and filho sat badly: being someone's son is arguably rigid — you do not stop being your father's son when he dies — yet it is plainly externally dependent. Under a single dependence test the case is unproblematic.

What is lost, stated plainly. Two things.

1. Criança and pessoa now carry the same type despite differing in rigidity. The type no longer records that one is anti-rigid.
2. Any export to a rigidity-aware ontology (UFO, gUFO) maps object one-to-many — to Kind or to Phase — and that decision must be made at export time rather than read off the stored type.

Neither cost is incurred today. If an export of that kind becomes a requirement, reinstating phase is an additive change: no existing assignment becomes wrong, and the affected lemmas are identifiable by inspection (§12.2).

No abstract type#

Abstract nouns do not need one. DUL's Abstract is not "abstract nouns" — it is value spaces: Region, TimeInterval, SpaceRegion, the space of natural numbers. Meanwhile dul:Object explicitly covers social and cognitive entities: SocialObject, InformationObject, Concept and Description all sit under Object. So lei, ideia, teoria, poema are Objects in DUL already. The lexicographer's "abstract noun" and DUL's Abstract are different notions that share a word.

The only residue is fact and proposition nouns (o fato de que…, a possibilidade), which are reified Situations. These take event. Minting a type for them would not pay for itself.

Mass nouns need no separate type either: água is anti-unity but a rigid sortal, and unity is not a typing axis here (§3).

If dot objects (livro = physical • information) later require it, add a boolean refinement on object — physical vs social-informational — not a base type. Deferred until co-predication actually bites.

5. Conceptual diagnostics#

The first question: conceptual or procedural?#

Does the lemma contribute a constituent to the proposition, or does it instruct the hearer how to process one? If the latter, go to §7 and stop.

object / role#

Both are thing-like, and one binary test separates them: external dependence.

	Type	Examples
not externally dependent	object	pessoa, cadeira, documento, criança, adulto
externally dependent	role	comprador, filho, presidente, proprietário

The dependence test, replacing constitutivity. The previous version asked "does the relation have to be invoked to make sense of the word?" — an intuition probe with no external anchor. Ask instead:

Is there an entity y, not a part or constituent of x, whose existence is required for x to be P?

Comprador → yes (a purchase, a seller). Filho → yes (a parent). Documento → no. Criança → no: being a child requires no second entity, so it is object, not role — even though an individual stops being one (§4, "Why rigidity is not a typing axis").

The old warning still applies and is sharpened by the new test: mere participation is not dependence. Documents get signed, filed and sent, but no particular event must exist for a document to be a document. If participation sufficed, every entity noun in the lexicon would be a role.

Two corpus diagnostics, for cases where the ontological test is hard to apply. Both operationalize Löbner's [R]:

1. Bare predicative: ?Ela é irmã vs Ela é médica. A +R concept resists the bare predicative without a recoverable relatum.
2. De-genitive reading: does de X give an argument (pai de João) or a modifier (carro de João)? Argument readings indicate +R.

These are annotatable and scorable, which means the role boundary can be validated rather than defended by introspection.

event / object in deverbal nominals#

A deverbal noun is an event by default and stays one. Nominalization is a constructional fact: it lets an event concept be referred to, counted, determined and modified like a thing, and none of that changes what the concept is. Duas destruições, a destruição foi total, aquela destruição are all the event under nominal packaging. Countability, determination, pluralization and adjectival modification are never evidence for object.

The one case that does diverge is a lemma with a product sense: a participant of the event — typically its Result or Product — with existence independent of the event, which persists after it ends.

Does the lemma have a sense denoting a thing that the event produced, rather than the event itself under some construal?

Three tests, all of which the product sense passes and the event sense fails:

1. Spatial location. A pintura está na parede. / *A destruição está na parede.
2. Physical or material predicates. A tradução tem 300 páginas. / *A destruição tem 300 páginas.
3. Persistence. The product exists once the event is over; the event does not.

Pintura, construção, tradução, invenção, criação, publicação pass. Destruição, aquecimento, chegada, corrida, viagem fail: they have no product, and their apparent "result" readings are the event's Result or Effect FE reached metonymically, which the frame layer records and the type layer must not.

Do not use Grimshaw's diagnostics here. Her complex-event / simple-event / result trichotomy separates nominals by argument structure: whether the noun takes obligatory arguments and aspectual modifiers. Simple event nominals (viagem, corrida, evento) fail those tests while remaining ontologically events, so the tests do not identify type. They are a useful description of nominal syntax and a misleading guide to this decision.

quality / value#

Ask whether the lemma names the dimension or a region within it. Altura / alto, temperatura / quente, beleza / bonito, cansaço / cansado are each a dimension/value pair, decided the same way every time.

No entity is presupposed by either. A temperature scale exists, and a point on it is nameable, whether or not any object is currently at that temperature — está quente hoje, said of the weather, names no entity at all. What links a value to the entity bearing it is the job of the Attributes schema (its Entity, Value and Attribute elements), not of the lemma's type. This is why there is no entity-flavoured value type, and never was a good case for one.

relation / role#

The tie is relation (paternidade, posse); the argument position named as such is role (pai, proprietário). Arity and converse evidence (pai-de ↔ filho-de, posse ↔ pertence-a) confirm a relation exists; per §4 the link between the two is dul:isDefinedIn, not a parameter.

A free-standing referential use of a role noun — pai picking out the individual who walked in, with no relation in play — is a determination-driven shift, not a second type. The lemma remains role (§1.4).

6. The state/attribute boundary is dissolved, not adjudicated#

The previous version distinguished .state from .attribute by whether a producing force or event was on record: cansaço (resultant) and postura (force-maintained) were .state; beleza and inteligência were .attribute. This distinction is removed. Both are now quality; their values are both value.

Three reasons.

It duplicates the frame. "Is a producing event on record?" is a question about the situation, and the frame layer is where situations are recorded. The document already concluded as much when @condition was retired: force- status was found to decide which conceptual schema a frame's meaning is read against (STATIVE vs Attributes plus a relation to the producing event), not which namespace the frame sits in. A fact that does not distinguish namespaces and is already carried by the frame's grounding does not need a type-level home — and now that the type sits on the lemma, a frame-level fact could not be recorded there even in principle.

It is the same asymmetry already accepted for destruição. If destruição stays event while its frames range over @agentive, @change and @stative, then quebrado stays value while its frame carries the fact that the value resulted from an event. Treating the two cases differently was inconsistent.

It imports a contested distinction with no reliable test. The state/ attribute line re-lexicalizes the stage-level vs individual-level predicate distinction (Carlson 1977; Kratzer 1995), which is gradient, frequently ambiguous within a single predicate, and which the semantics literature declines to treat as an ontological sort. It was the lowest-agreement decision in the scheme, and the agreement problem was structural rather than a matter of insufficient guidance.

Consequences: postura, cansaço → quality. Cansado, quebrado, sentado → value. Anti-rigidity and causal origin are read off the frame.

7. Procedural types#

Nine types. Procedural lemmas take no DUL reference and are exempt from the class-compatibility check (§10).

Type	Encodes	Examples
reference	how to resolve a referent — deictic, anaphoric, definiteness	o, este, ele, quem, aqui, hoje, agora
quantification	how to quantify over a domain	todo, cada, algum, vários, ambos, sempre
polarity	assert or deny instantiation	sim, não, nunca, jamais, nada, nenhum, sem
degree	how to locate a value relative to a standard	muito, pouco, mais, menos, bastante, demais, tão
modality	epistemic or deontic status of the proposition, with no scale named	talvez; modal auxiliaries (poder, dever, ter que); free mood markers
connection	inferential or discourse relation between units	contudo, portanto, porém, entretanto, pois, se
focus	information-structural instruction	só, também, até, mesmo, inclusive
predication	how to build the predication; contributes no content	copulas (ser, estar), auxiliaries, support verbs, case-marking de/a/por
interaction	how to manage the exchange — appeal, stance display, information source, hedging	né, hein, viu, tá, ué, nossa, olha, tipo, assim, diz que

The nine types are organized by functional category, not taken from Relevance Theory — see "Where these nine types come from" below, which also records where the inventory departs from RT and what it leaves out.

Notes on the boundaries#

degree vs value. The criterion is whether the lemma names a region or instructs where to locate one relative to a standard. Quente, morno, azul name regions → value. Muito, pouco, bastante do not name a region; they locate one against a contextually supplied standard → degree. This is why muito is no longer typed as a value on a "grau" dimension.

Prepositions are not a single case. ADP splits across both sorts: contentful spatial and temporal adpositions are relation, governed case-markers are predication, and some are polarity or connection. See §8.2.

modality vs value: does the lemma name a scale? Epistemic adverbs with a lexicalized dimension are value, not modality. Provavelmente, certamente, possivelmente and seguramente each name a region on the probability/certainty dimension (probabilidade, certeza), have a gradable adjectival counterpart (provável, certo), and stand to it exactly as belamente stands to beleza — which §8.1 types value. Three tests separate them from modality:

Test	value — provavelmente	modality — talvez
degree modification	muito provavelmente ✓	*muito talvez ✗
lexicalized dimension and adjective	probabilidade, provável ✓	none ✗
embeds under negation	não é provável que… ✓	*não talvez… ✗
governs mood	no	yes — subjunctive

The first test is decisive on its own: degree modifies value by definition, so anything muito can stack on is a value. Talvez fails all three and governs the subjunctive, which is an instruction to the grammar rather than a region on a scale. Modal auxiliaries form a third group with no scale at all and stay in modality.

The copula and support verbs are predication. They instruct how to assemble a predication and contribute no concept of their own. This removes two long-standing hard cases at once.

interaction is about the exchange, not about the proposition. It covers interjections (ué, nossa, ai), interactional and tag particles (né, hein, viu, tá), attention markers (olha, escuta), hedges and approximators (tipo, assim), and quotative or evidential markers (diz que, dizem que). What unites them is that they manage the interaction — appealing to the hearer, displaying a stance, flagging where information came from — rather than contributing to or operating on the proposition expressed.

Three boundaries need holding:

• vs connection. A connective relates two propositions inferentially (portanto, contudo); an interactional particle relates the utterance to the hearer (né seeks confirmation, viu secures uptake). Test: does removing it leave an inferential gap between two units, or only a blunter utterance?
• vs modality. modality marks the speaker's own commitment to the proposition (talvez, poder); evidentials mark where the information came from (diz que). The two are adjacent and some frameworks merge them; FNBr keeps evidentials in interaction because they invoke the hearer's assessment of the source rather than stating the speaker's confidence (§12.11).
• vs focus. Só, até, mesmo operate on the proposition's information structure and are focus. Tipo and assim as hedges do not restructure the proposition; they signal approximation in the exchange.

interaction lemmas evoke pragmatic frames like any other lemma evokes frames. They are exempt from the class-compatibility check of §10 along with the other procedural types, which is why no pragmatic namespace appears in the admit-set table.

Where these nine types come from#

The conceptual/procedural split is Relevance Theory's (§2). The nine subtypes are not, and this document should not be read as claiming otherwise: no RT publication proposes this inventory.

RT subdivides procedural meaning along two axes, neither of which is used here. The first is the target of the constraint — explicature, higher-level explicature, or implicature (Blakemore 1987; Wilson & Sperber 1993), with Blakemore further separating contextual implication (so), strengthening (after all) and contradiction (but). The second is the cognitive subsystem the item triggers (Wilson 2011, 2016): emotion recognition, epistemic vigilance, social cognition, intention recognition.

The inventory here is organized by functional category instead, which makes it an application of Escandell-Vidal & Leonetti's thesis that the semantics of functional categories is procedural. Their own enumeration — discourse markers, sentence-mood marks, quotative and evidential particles, intonation, verb tenses and moods, definite determiners and pronouns, deictic and focusing adverbs, information-structure devices — is a list, not a typology. The nine types are a coarsening of that list into classes an annotator can assign.

Type	Nearest published support	Status
reference	Wilson & Sperber 1993 on pronouns; Escandell-Vidal & Leonetti on definites and deictics	RT-aligned
connection	Blakemore's core case	RT-aligned
focus	focusing adverbs in Escandell-Vidal & Leonetti; Iten on concessives and even	RT-aligned
interaction	Wharton on interjections; Curcó on Spanish particles; quotative and evidential particles in Escandell-Vidal & Leonetti	RT-aligned
predication	the functional-category thesis	derived, not RT
quantification	only definite determiners are standardly procedural in RT	departure
polarity	classical RT gives não/not a logical entry, i.e. conceptual	departure
degree	degree semantics (Kennedy; and Gärdenfors for the scale)	outside RT
modality	RT treats only mood and modal particles as procedural	RT-aligned after the narrowing in §7

Departures, and why they are taken#

quantification and polarity. Classical RT (Sperber & Wilson 1986, ch. 2) treats logical words as concepts with logical entries. That claim is about how an item enters inference; the criterion used here is different and narrower — does the lemma name something the conceptualization inventory can hold? Não, todo and nenhum have no dul: referent under any construal, and typing them conceptually would require inventing one. The departure is taken for groundability, and is deliberate.

degree. Not an RT category at all. The "region located against a standard" criterion comes from degree semantics, and the dimension/region framing from Gärdenfors (§3). It is listed as procedural because it instructs rather than names, which is the operational test of §2, not because RT puts it there.

modality was a conflict and has been resolved. Wilson & Sperber (1993) and Ifantidou (1993, 2001) treat sentence adverbials such as certamente and provavelmente as conceptual, contributing to higher-level explicatures; Papafragou (2000) treats modal verbs as conceptual. In RT only mood and modal particles are procedural. An earlier draft of this document put the scalar adverbs in modality, which contradicted that source and also failed this document's own test in §8.1: provavelmente names a region on a likelihood dimension rather than instructing where to locate one, which is the criterion that sends quente to value and muito to degree.

The scalar adverbs are now value and modality is narrowed to talvez, modal auxiliaries and free mood markers (§7, "modality vs value"). The row is RT-aligned as a result. One residue remains: RT would also call the modal auxiliaries conceptual, following Papafragou. They are kept procedural because no conceptual type fits them — they name no scale, no event and no relation — and because predication would misdescribe them as contentless. This is a smaller departure than the one it replaces, and it is deliberate.

What this inventory still lacks#

Sentence mood and type (declarative, interrogative, imperative) has no type of its own. It is currently spread across reference (quem) and modality, and where mood is carried by free words rather than by inflection or intonation, those words land in interaction. Whether mood deserves its own type depends on how much of it FNBr lexicalizes; a mood realized by verb morphology or intonation is not a lemma and is out of scope regardless.

Evidential and quotative particles (diz que) and interjections and interactional particles (né, ué, hein, tipo) are covered by interaction, added for FNBr's pragmatic frames. The boundary with modality is live — see §12.11.

Tense and aspect are procedural on most RT accounts but are out of scope here for a different reason: they are bound morphology, not lemmas.

One RT axis considered and declined#

Adding target ∈ {explicature, higher-level explicature, implicature} as an orthogonal attribute would let the scheme cite Wilson & Sperber (1993) directly. It is not adopted. The target of a procedural constraint varies with use — the same connective can strengthen or contradict — so the attribute would be a per-token judgement recorded on a per-lemma row, and nothing in §10 consumes it. This document has removed judgement-based axes elsewhere on the same grounds (§4 on rigidity, §6 on state/attribute), and adding one here would be inconsistent.

8. Parts of speech the type layer cuts across#

Some UPOS tags collect lemmas with almost nothing semantically in common. The type layer is expected to cut across them, and two tags in particular need worked guidance: ADV and ADP. Neither is a semantic class, and in both cases the members distribute across the conceptual and procedural sorts.

8.1 Adverbs#

"Adverb" is a part-of-speech wastebasket, not a semantic class — it collects manner, place, time, frequency, degree, epistemic and connective words with almost nothing in common. The type layer is expected to cut across it. Under the two-sort vocabulary, the class splits cleanly, with most members turning out to be procedural.

Adverb class	Encodes	Type
Manner (belamente, rapidamente)	a region on a quality dimension, predicated of an event	value
Time extent (recentemente)	a region on a temporal-distance dimension	value
Frequency, graded (raramente, frequentemente)	a region on the frequency dimension	value
Degree / intensity (muito, pouco, bastante)	locate a value against a standard	degree
Universal frequency (sempre)	quantification over times	quantification
Epistemic, scalar (provavelmente, certamente, possivelmente)	a region on the probability/certainty dimension	value
Epistemic, non-scalar (talvez)	epistemic status, no scale named; governs subjunctive	modality
Epistemic stance as a concept (dúvida, incerteza, certeza)	the dimension itself	quality
Place and deictic time (aqui, hoje, agora)	resolution instruction	reference
Conjunctive (contudo, portanto)	discourse relation	connection
Affirmation / negation (sim, não, nunca, nada, nenhum)	polarity	polarity

Manner adverbs are values, not dimensions. Belamente stands to beleza exactly as bonito does — a region on the same dimension, predicated of an event rather than an entity. What the value is predicated of is a frame-level fact, not a type-level one.

Deictics are procedural, not relations. Aqui and hoje were previously typed .relation on the grounds that they locate a figure relative to a ground. They encode an instruction to resolve a referent against the utterance situation, which is the textbook case of procedural encoding. Their image-schematic character is carried by what their frames ground in (FIGURE_GROUND, DEICTIC), where it always was.

Two diagnostics retained from the previous version

Zero is polarity, not a low value. Nada and nenhum look like values on the quantity dimension, parallel to pouco. But a low value presupposes existence — pouco asserts some quantity, just a small one — while nada and nenhum deny existence outright (¬∃x). The same line separates raramente (a low but positive frequency, still asserting the event occurred) from nunca (a negative existential over time points). A zero reading on any dimension negates that dimension's existential presupposition rather than measuring a small amount of it.

The diagnostic survives intact and its conclusion is now better supported. Previously it delivered nada to .relation, which required claiming that a polarity operator is a tie between entities — defensible only under a strained construal, and the weakest point in the old specification. It now delivers nada to polarity, which is simply true of it.

The polarity relation is not the act. Não as propositional polarity is polarity; the act of affirming or denying (afirmar, negar) is an event in an @agentive frame. Negação carries both readings under one form and one POS, and is split into two lemmas accordingly (§1.4).

8.2 Prepositions and adpositions#

ADP is one UPOS covering lemmas that fall on both sides of the conceptual/procedural line, and the split does not follow the tag at all. A preposition may name a relation, mark a case, express polarity or carry a discourse connection.

Use	Examples	Type
Spatial or temporal configuration	sobre, sob, entre, contra, durante, após, dentro de, ao lado de, antes de	relation
Privative	sem	polarity
Causal, concessive, conditional	por causa de, apesar de, devido a, em vez de	connection
Governed by the predicate (case-marking)	de in gostar de, a in obedecer a, em in acreditar em, com in contar com	predication
Marking a role the frame already supplies	por (passive agent), com (instrument), para (beneficiary)	predication

Three diagnostics

1. Contrastive substitution. Holding the governing head constant, can another preposition replace it with a truth-conditional difference? O livro está sobre / sob a mesa — yes, so relation. *Gostar em — no, so predication.
2. Questionability. Can the PP answer a wh-question with the preposition inside the answer? Onde? — Sobre a mesa. Governed prepositions fail this: there is no question whose answer is de in gostar de.
3. Government. Is the preposition listed in the valence of the head that selects it? If yes, it is predication regardless of what it means elsewhere.

Diagnostic 3 overrides the other two when they conflict, because government is observable in the valence data while the other two rest on judgement.

Governed prepositions are markers, not frame-evoking lemmas

A case-marking preposition contributes no concept and should not be given a frame. Its place is the valence pattern, as the marker of an argument position — the role synsem:marker plays in OntoLex-Lemon. The predication type records that the lemma is real and type-bearing but conceptually empty; which head governs it is recorded at the valence layer, not by a frame link. Creating frames for governed prepositions duplicates the valence pattern and populates the lexicon with frames no annotator can define.

Applying the rule: prepositional splits

Adpositions are the largest application of the rule in §1.4 — one lemma per (text, POS, type) — and the reason it was adopted. A single de row would carry a declared type that is wrong roughly half the time it is used, which no default could repair: neither reading is exceptional and both are extremely frequent.

Splitting stays bounded because readings that share a type do not split. De has at least six distinguishable readings but only two types, so it yields two lemmas, not six; genitive, material and partitive are LUs of the same predication lemma because they are all markers.

Lemma	Type	Covers	Examples
de₀	predication	government; genitive; material; partitive	gostar de música; o carro de João; mesa de madeira; um copo de água
de₁	relation	source and origin, spatial or temporal	veio de São Paulo; de segunda a sexta
a₀	predication	government; dative and recipient marking	obedecer a uma ordem; dei o livro a João
a₁	relation	goal, direction, temporal location	fui a Roma; às três horas
em₀	predication	government	acreditar em algo; pensar em você
em₁	relation	containment, location, temporal location	na mesa; em 2020; no Brasil
por₀	predication	passive agent; government; price marking	destruído por João; optar por outro; comprei por dez reais
por₁	relation	path, duration, cause	passou por São Paulo; por três horas; preso por roubo
com₀	predication	government; instrument and manner marking	contar com você; cortou com a faca; com cuidado
com₁	relation	comitative, accompaniment	foi com Maria; café com leite
para₀	predication	government; beneficiary and recipient marking	olhar para a janela; comprei um livro para João
para₁	relation	goal, direction	fui para Roma; para o norte
para₂	connection	purpose, linking two events	saiu para comprar pão
sobre₀	relation	superposition	o livro está sobre a mesa
sobre₁	predication	topic marking	um livro sobre linguística
sem₀	polarity	privative — no split, one type covers every reading	sem açúcar; saiu sem pagar

Two things the table shows. Sobre splits even though it looks like a paradigm case of a contentful preposition, so the split is not a property of "grammatical" adpositions only. Sem does not split at all, so the split is not automatic either: it is driven by type divergence and nothing else.

The cost of splitting, and how to pay it

Splitting breaks the disambiguation story of §1.2. There, form lookup returns candidates and UPOS filters them — which works for colarᵥ / colarₙ because the candidates differ in POS. de₀ and de₁ share both text and POS, so UPOS filters nothing. Disambiguation must come from the syntactic context instead.

This is workable because the needed evidence is in the UD parse already:

1. Government. If the adposition's head is a predicate that selects it — the case dependent of an obl/obj whose head lists that preposition in its valence — choose the predication lemma. This is diagnostic 3 above, and it decides most tokens.
2. Complement semantics. A place or time complement favours the relation lemma; an infinitival complement under para forces para₂.
3. Default. Where neither fires, prefer the predication lemma. It is the more frequent reading for de, a, em, por, com and para, and a marker reading is the less damaging error: it attributes no content that is not there.

The honest summary is that splitting moves work from the lexicon to the parser. That is the right direction — the lexicon records a stable fact and the parser resolves a contextual one — but it is a real cost and should be measured on the adposition tokens specifically, not averaged into overall lemma-resolution accuracy.

The rule is not about part of speech

Nothing here is specific to adpositions. The same rule produces negação₀ / negação₁ and pintura₀ / pintura₁ in the content lexicon (§1.4). Adpositions are simply where it fires most often, because a small set of very frequent forms each span the conceptual/procedural line.

The practical consequence is that the split count is concentrated: a handful of function words account for most of it, and a lemma table with a few dozen split forms is not evidence of a problem with the type inventory (§12.7).

Two structural notes

Complex prepositions are MWE lemmas. Dentro de, ao lado de, por causa de, apesar de take entry_type = mwe with pos nullable (§1.2). Their type follows the table above and is often connection rather than relation, which is a further reason not to read the type off the tag.

The type cuts across ADP/SCONJ as well as within ADP. Por causa de (ADP) and porque (SCONJ) are both connection; para + infinitive sits on the boundary. This is the same phenomenon as §8.1 and the reason this section exists: the tag predicts the type in neither direction.

Consequence for the check in §10

The structural check requires a relation lemma's frames to have FEs lexicalized by role lemmas. Spatial prepositions systematically fail this: the FEs of a locative frame are Figure and Ground, and Portuguese has no role lemmas naming them. The check is therefore scoped in §10 to relational frames whose FEs have role lexicalizations; without that scope it would flag every contentful preposition in the lexicon.

9. Why coarse: the three layers#

The asymmetry is the most important property of this design, and with the type on the lemma it reads as three layers rather than two:

Layer	Grain	Carries
lemma	coarse identity	the type — what kind of thing the lexeme names
LU	sense	the lemma–frame pairing and a sense description — no type
frame	situation	the namespace, the participant structure, the grounding schema

Hence:

• Destruição is one lemma typed event. Its LUs may sit in an @agentive frame (destruction as cause), a @change frame (destruction as change of state) and a @stative frame (destruction as resulting condition) — without contradiction and without splitting the lemma. The type is declared once and inherited by all three.
• Aquecer is one lemma typed event, with LUs in Causar_mudança_de_temperatura and Mudança_de_temperatura. Same type, different frames. This is the ordinary case, and it is exactly what moving the type to the lemma is meant to capture: the fact was never per-LU.
• value behaves the same way across @attribute, @stative and @entity. Americano names a region on the origem dimension, full stop, regardless of what it is predicated of: a person (um americano), a thing (um carro americano) or an event (o ataque americano). What changes is only which frame each LU evokes. frm_people_by_origin is an @entity frame built over value lemmas, and it is not an exception: the live data has dozens more (frm_people_by_religion, frm_people_by_vocation, frm_wealthiness, frm_expertise, and around twenty-five others).
• POS differences split the lemma, not the LU. Dependenteₙ ("meu dependente") is role; dependenteₐ ("ele é dependente") is value. These are two lemmas by §1.2, each with its own type, each with its own LUs.
• Type divergence splits the lemma too. Where one form under one POS carries two types, it is two lemmas by §1.4. Nothing in the LU table records a type, so no divergence can be hidden there.

Determination-driven type shifts change neither. Löbner shows that congruent determination and marked type shifts are regular and productive: ele é pai shifts a functional concept to a sortal predicate without changing the underlying concept. Shifts of this kind are not type ambiguity, do not spawn LUs, and do not split the lemma. The lemma's declared type stands — which is exactly what commitment 3 in the preamble says.

10. What this layer validates, and what it cannot#

All checks run on lemma.type_id. There is no second source to reconcile: an LU's type is its lemma's type, always.

The type licenses a class-compatibility check: does the frame's namespace admit this type at all?

Namespace	Admits
@agentive, @change, @phenomenon, @process, @undergoing	event
@experiential	event
@stative	event (the maintaining verbs — permanecer, manter — are eventive lemmas like any other verb evoking an eventive frame), quality, value
@attribute	quality, value
@entity	object, role, value
@relational	relation, role
—	procedural types are exempt

A second, structural check is now available, which the parameter scheme could not express:

For every relation lemma, the frames its LUs evoke should have FEs lexicalized by role lemmas, and those roles should be isDefinedIn that same frame-as-Relation.

Scope. The check applies only to relational frames whose FEs actually have role lexicalizations. Spatial and temporal frames are excluded: their FEs are Figure and Ground, which no Portuguese lemma names as a role, so every contentful preposition would otherwise be flagged (§8.2).

A relation lemma whose frames have no role lexicalizations is either mistyped or sitting in a frame missing its participant LUs. A role lemma whose frames are not Relations is a typing error. Neither was detectable before, because .entity<relation> and .relation carried no structural link to each other.

A third check becomes available only because the type sits on the lemma:

A lemma whose LUs span frames with incompatible admit-sets is a candidate for a split (§1.4).

This is the useful diagnostic that the previous per-LU design could not produce at all, since every LU carried its own type and no tension between them was ever visible.

What the checks do:

• Flag cross-class category errors: a quality lemma in an eventive frame, an object lemma in a stative frame, a relation lemma in an agentive frame. Automatable, high-precision. Per commitment 3, a flag is a marked combination for review, not a verdict.
• Silent on within-class distinctions: @agentive vs @change vs @phenomenon all admit event, so the check says nothing about which is right. Both aquecer LUs pass.
• Silent on wrong-sense errors: only an annotator working against a real sentence can choose which frame an occurrence belongs to.
• Does not assign types automatically. Which of the six ontological types, or whether a lemma is procedural at all, is annotator judgement guided by §5–§8.

11. Migration#

11.1 From the parameter scheme#

The nine LU-scope namespace values are replaced by fifteen — six ontological and nine procedural — now scoped to lemma. The dotted notation remains a documentation convention: nothing in the schema stores .event or .role as text; each lemma carries an idNamespace link to one namespace value, the same device frames and FEs use.

destruição.event     comprador.role      parentesco.relation
cadeira.object       cansaço.quality     cansado.value
criança.object       não.polarity        portanto.connection

Previous	Now	Note
.event	event	unchanged
.entity	object	renamed
.entity<event>	role	parameter retired; the relatum's sort is carried by the frame
.entity<relation>	role	same
.attribute	quality	renamed
.state	quality	boundary dissolved (§6)
.attribute<value>	value, or degree / modality / quantification	scalar adverbs reassigned (§8)
.state<value>	value	boundary dissolved (§6)
.relation (parentesco, posse)	relation	grounding sharpened to dul:Relation
.relation (não, contudo, aqui)	polarity / connection / reference	moved to the procedural sort (§7)

Anti-rigid intrinsic sortals (criança, adulto, novato) take object. A phase type was specified for them in an intermediate draft and then folded into object; if any data was assigned under it, remap phase → object (§4, "Why rigidity is not a typing axis").

11.2 From LU-scope to lemma-scope#

The namespace scope moves from lu to lemma, and the LU table returns to its FrameNet shape exactly: a lemma × frame association with a sense description and nothing else.

Steps:

1. Add type_id to lemma, nullable during migration. Do not add a type column to lu.
2. For each lemma, collect the types of its existing LUs. Where all agree — the expected majority — write that type to the lemma and drop it from the LUs.
3. Where they disagree, the lemma splits (§1.4). Create one lemma row per distinct type, assign homograph_idx in descending frequency order so the dominant reading keeps 0, and repoint each LU at the row matching its type. A POS difference means the rows were already two lemmas and nothing needs deciding.
4. Regenerate the form paradigms for the new rows. Split lemmas share a paradigm, so this is a copy, not a re-derivation — but the form table is keyed on lemma_id and will otherwise leave the new rows formless.
5. Make lemma.type_id non-null and drop any LU-level type column.
6. Record the split inventory. It is the input to §12.7 and to the parser work in §12.9, and it is far more useful captured during migration than reconstructed afterwards.

Step 3 is the only one requiring judgement, and step 4 is the one most easily forgotten. Neither is reversible cheaply once annotation resumes.

11.3 Why parameters were retired#

The <event> / <relation> parameter recorded which relational concept individuated an entity. That fact is already carried by the frame: comprador is a role in a Commerce_buy frame, filho in a Kinship frame. Recording the relatum's sort on the lexical entry duplicated the frame link.

The entity-base vs relation-base debate documented in the previous version is thereby closed rather than decided. It asked whether the coarse identity or the relational concept should occupy the base slot. With role grounded in dul:Role and relation in dul:Relation, both are base types and both are coarse; the DnS pattern supplies the link between them, including the converse handling that was the main argument for relation-base.

The <value> parameter is likewise retired: quality and value are sibling base types, which is what the dimension/region distinction always was.

12. Open questions#

Recorded so they are not rediscovered as objections.

1. Validation of role. What inter-annotator agreement does the dependence test achieve on Portuguese data? The two corpus diagnostics in §5 give an external anchor; the number has not yet been measured.
2. Rigidity is not recorded. Criança and pessoa share the object type despite differing in rigidity (§4). This costs nothing internally, but an export to UFO or gUFO would have to decide Kind vs Phase at export time. If that export becomes a requirement, reinstate phase — an additive change that invalidates no existing assignment. A candidate list can be generated by querying object lemmas whose frames are @entity frames built over life-stage, career-stage or status dimensions.
3. [U] is not used. Löbner's uniqueness feature distinguishes filho (relational) from pai (functional) and predicts determination and bridging behaviour. FNBr currently collapses both to role. Whether [U] earns a column is an open question.
4. Dot objects. object does not represent the dual identity of livro (physical • information) that co-predication exposes. Deferred, not solved.
5. POS × type correlation. How much independent information does the type layer carry over UPOS? One contingency table answers it, and the answer determines how strongly §1.5's claim can be stated.
6. The two-sort split has a boundary. Some lemmas plausibly encode both conceptually and procedurally (sempre, só). The current rule assigns one type; whether that is adequate is untested.
7. Split count. How many lemmas split under §1.4, and how concentrated is the count? A few dozen forms, dominated by adpositions, is the expected shape (§8.2). If splits are instead spread thinly across the content lexicon, the type inventory is drawing a line the lexicon does not observe, and the inventory — not the rule — is what needs revisiting.
8. Relatedness between split lemmas is unrecorded. Negação act/operator and pintura activity/canvas are now separate rows with no link. Where the relation matters it should be added explicitly at the lexical layer (OntoLex-Lemon vartrans). Whether it matters enough to build is open; nothing in the current checks needs it.
9. Adposition disambiguation accuracy. Splitting de, a, em, por, com, para and sobre by type (§8.2) moves their disambiguation to the parser, where UPOS cannot help because the candidates share it. Accuracy on adposition tokens should be measured separately rather than averaged into overall lemma resolution, and the government heuristic evaluated on its own. This is the largest single risk the design carries: if the heuristic underperforms badly, the alternatives are a richer valence-driven disambiguation or a coarser type inventory for adpositions — not a return to per-LU typing, which would reintroduce the annotation burden the rule exists to remove.
10. The modality / value line needs validating, not deciding. The decision is taken (§7): scalar epistemic adverbs are value, talvez and the modal auxiliaries are modality, and the criterion is whether the lemma names a lexicalized dimension with a gradable counterpart. What is untested is coverage — whether every PB epistemic adverb falls cleanly on one side. Possivelmente and eventualmente are the likely stress cases, and eventualmente may not be epistemic at all in PB. Run the three tests over a closed list of epistemic adverbs before annotation opens.
11. The modality / interaction boundary for evidentials. Diz que is currently interaction on the grounds that it points at an information source rather than stating speaker confidence. Some frameworks treat evidentiality as a species of epistemic modality, which would put it with modality. Now that modality is narrowed to non-scalar items, the two are closer than before. A pilot on pragmatic-frame data will settle it faster than argument will.
12. interaction has not been tested at scale. It was added for FNBr's pragmatic frames and its three boundaries — against connection, modality and focus (§7) — are stated but unvalidated. PB interactional particles are frequent and often multifunctional (tá as tag, as agreement token, as imperative of estar), so this type is the most likely source of new §1.4 splits.