Type classification for lemmas

Each lexical lemma is assigned one type under this model; punctuation and symbols are outside its scope (§6), and untyped records require review rather than a guessed value. 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 type-classification decision; it does not decide the lemma's LU senses or their allocations.

A lemma here is a lexeme: a citation form paired with a part of speech, not a bare string. A lexical unit (LU) records a reading of a lemma in a frame, carrying a sense description. Lemmas and frames have a many-to-many relationship. Shared lemma/frame IDs are duplicate-candidate evidence, not proof that two LU records represent the same reading. 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.

  1. The type is a property of the lemma — a lexeme identified by language, text, POS and any homograph distinction — not of a word form or a referent. POS already does the construal-fixing work: colar/VERB and colar/NOUN are different lexemes with different meanings 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 follows the grounding ontology: DOLCE is a Descriptive Ontology, an inventory of particulars as conceptualized by natural language, explicitly neutral on metaphysics. Typing a lemma against DUL types 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, since a function needs an argument. 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 — not a descriptive category of Portuguese, and not a metaphysical claim.

Because the type is defeasible (commitment 3), and because it does not restrict which frames a lemma's LUs evoke (§8), a type that looks unusual for a frame is not by itself an error.

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

1.1 Why retain POS at all#

The theoretical case against parts of speech as a semantic category is strong: syntactic categories are language-particular and construction-derived (Croft, Radical Construction Grammar), and do not translate across languages as descriptive categories (Haspelmath 2010). This is why the type layer exists and why it is not a POS layer.

POS is nonetheless retained: it is load-bearing for lemmatization and morphological analysis, MWE and terminology identification, corpus query (word sketches are defined over POS patterns), and interoperability with UD, LMF and OntoLex-Lemon/LexInfo.

The architecture follows OntoLex-Lemon: 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 part of lexical-entry identity, together with language and any homograph distinction. The following is a logical identity model, not a claim about physical column names or implemented database constraints:

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. 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 would resolve colo to a single colar row and make the noun sense reachable from a verb form — a lemma that does not exist.

Two consequences:

  • Form lookup is set-valued. form.text → {lemma_id} returns candidates. POS is what filters them, but POS lives on the token, not on the form string: parsing the sentence assigns the token a UPOS, and matching that against the POS carried by each candidate lemma picks the right one. Colo is itself ambiguous off any list — it is a form of colar/VERB and also a noun in its own right (colo/NOUN, "lap") — so the string alone does not resolve it. A reliable token POS can exclude the noun/verb mismatch, but homographs and readings sharing that POS still require disambiguation.
  • 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: 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 — that is what LUs are for (§1.4).

FNBr's entry convention uses pos = X for MWE entries, with entry_type ∈ {word, mwe, affix} distinguishing entry kinds. This is a local lexicon convention, not a claim that UD requires X for every MWE: an expression can have a syntactic category while its components receive individual POS tags. The expression's conceptual/procedural type is a separate question.

X, not null. A null POS propagates: view_lu derives an LU's display name as concat(lemma.name, udpos.suffix), and concat() returns null when any argument is — so a null POS silently blanks the rendered name. X carries a real suffix, so dar um jeito renders as dar um jeito.x. Every lemma carries a POS; a lemma arriving without one is assigned X rather than left null under the entry convention. The nullable storage field means existing missing-POS records still need validation; this documentation does not repair them or authorize replacing their identity automatically.

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. Colarᔄ and colarₙ are separate lemmas with separate types; so are dependenteₙ (role) and dependenteₐ (value). Putting the type on the lemma–frame pairing instead would record, on every LU, a fact constant across all of a lemma's LUs.

The type is available before frame assignment. A parser can type a token as soon as it resolves the lemma, with no frame in hand — which is precisely when a type is most useful as a filter.

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

1.4 Type divergence and lemma identity#

Within a language and lexical identity, established readings of different types require separate lemma identities. Readings sharing a type can remain under one lemma and be distinguished as LUs. Genuine homonyms may remain separate even when text, POS and type coincide.

One type per lemma; type equality alone does not establish lemma identity.

This is the only mechanism for type divergence in the specification — there is no LU-level type and no override.

  • 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.
  • Stored assignments are reviewable. The intended single-valued type does not make the data correct by construction. Missing types, misallocated LUs, mixed readings and rule limitations must be distinguished before splitting. Namespace diversity alone does not authorize a lemma split.

The split is bounded by type, not by sense count: de has at least six distinguishable readings but only two types, so it yields two lemmas, not six (§6.2).

Three case shapes are known:

  • Cross-sort polysemy. Negação names the act of denying (event, in an @agentive frame) and the propositional operator (polarity) → negação₀ event, negação₁ polarity.
  • Product nominalization. Pintura as the activity is event; as the canvas on the wall it is object. Same for construção, tradução, invenção, criação, publicação — deverbal nominals whose event produces an artifact that outlives it. Not all deverbal nominals: see §4.
  • Adpositions. De, a, em, por, com, para — the largest group, worked through in §6.2.

Not for:

  • Construal shifts. A lemma can be referred to, or otherwise used, in a way that leaves part of its normal profile inactive without the stored type changing — the type is a fact about the lemma; the construal is a fact about the occurrence. Pai used to pick out an individual with no relation in play is still the role lemma: the relation is simply not invoked by that use. Löbner's determiner-triggered readings are the best-studied case, but the mechanism is not limited to determiners — see the next bullet. Per commitment 3, a construal shift is defeasance in use, not a second type — and, since nothing about the lemma itself changed, not a second lemma either.
  • Substantivized adjectives used elliptically. O bonito, um americano pick out an entity the adjective characterizes — a person, a car, whatever the covert head recovers as — via the construction that lets any adjective head an NP under a determiner with its head noun elided. This is a construal shift like the one above: bonito and americano stay value, and which entity the value is predicated of is a frame-level fact (§7), not a lemma-level one. The test: can the use be regenerated as "[some contextually recoverable sortal] that is ADJ," with the adjective keeping its ordinary value? If the covert head is freely recoverable and swappable by context, it is a construal shift and the adjective lemma is untouched. If the "noun" reading is fixed regardless of context and carries content the adjective does not — typically a specific institutional or relational profile — the use has lexicalized, and §1.2's POS split applies instead: dependenteₙ ("meu dependente", a role — someone for whom a specific provider must exist) is a different lemma from dependenteₐ ("ele Ă© dependente", value), not a construal shift of it (§7).
  • Polysemy within a type. Readings can share a typed lemma while requiring different LU analyses. Genitive, material and partitive uses of de need no LUs: their relation is evoked by the head or by a construction (§6.2). Neither multiple frames nor a screening conflict alone establishes type divergence.

Splitting is the same operation as homonymy handling, deliberately: manga fruit/sleeve and de₀/de₁ can both require separate lexical identities. The reason for separation and relatedness of the entries still need evidence; type classification does not erase a genuine homonym distinction.

A split records no relation between the two lemmas. Where relatedness matters — negação act/operator, pintura activity/canvas — it belongs at the lexical layer as an explicit link (OntoLex-Lemon's vartrans module), not as a shared row.

1.5 POS and type are not orthogonal#

Deverbal nominalizations and participles show strong type/POS correlation, because category is assigned in a structure that also determines event structure. The claim is narrower: the type is not derivable from POS and not reducible to it. How much independent information it carries is answerable by a POS × type contingency table, since both columns 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 classifiable as objects, events or qualities — the "world thing" is computed in context, if at all.

Relevance Theory supplies the distinction. 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, reference resolution, or 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 (§3) yes
procedural an instruction for processing no no

Procedural lemmas carry no DUL reference — there is nothing in the conceptualization inventory for them to reference, because they name an operation on a conceptualized particular, not the particular itself.

3. The ontological types#

Six types. Each grounds in a DUL class. There are no parameters.

Type DUL grounding Criterion Examples
event dul:Event occurs or happens: it exists in time, has (or can have) temporal parts, and does not persist unchanged destruição, aquecer, correr, quebrar, derreter, existir
object dul:Object a sortal — it supplies a criterion of identity — whose instances need no other entity to exist cadeira, água, pessoa, documento, lei, ideia, criança, adulto
role dul:Role ⊑ dul:Concept a sortal whose instances are externally dependent: some other entity, not a part of them, must exist comprador, jogador, cliente, filho, pai, proprietário, metade
relation dul:Relation ⊑ dul:Description a reified relation type — a Description that defines the argument positions (roles) it relates parentesco, paternidade, posse, casamento, adjacĂȘncia, anterioridade
quality dul:Quality a dimension — a gradable or categorial axis — that inheres in some bearer cor, tamanho, beleza, inteligĂȘncia, temperatura, cansaço, postura
value dul:Region ⊑ dul:Abstract a region or point within a quality's dimension, nameable without presupposing any bearer azul, quente, bonito, inteligente, cansado, quebrado, sentado, belamente

The criteria draw on rigidity, sortality, external dependence, and the dimension/region distinction — defined in full in Appendix §A.1.

The reference principle#

The lemma references a conceptualization, never a particular and never a formal object:

  • 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.
  • adjacĂȘncia and anterioridade reference the commonsense dul:Relation, not the formal counterpart under dul:FormalEntity.

This is what "semantics by reference" means for a lexicon: 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:

FNBr object DUL DnS relation
Relation 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 — not two unrelated primitives that happen to sound alike.

The instance layer, where 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 — the same coarse-identity/fine-situation asymmetry the whole model runs on (§7).

Role definition is not lexicalization home. A role can be defined in a relation while its participant-denoting LU evokes an Entity frame. FNBr requires this for agentive names. Record the participant association separately; see FrameNet patterns.

Converse pairs require semantic evidence. Shared isDefinedIn links identify roles in a common relation, but do not alone establish which roles are converses.

Rigidity is not used as a typing axis, and abstract nouns do not get a type of their own — both are informational notes rather than operative rules; see Appendix §A.2 and §A.3.

4. Conceptual diagnostics#

The first question: conceptual or procedural?#

Does the lemma contribute a constituent to the proposition, or instruct the hearer how to process one? If the latter, go to §5 and stop.

object / role#

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

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

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 (Appendix §A.2).

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] (Appendix §A.1):

  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 JoĂŁo) or a modifier (carro de JoĂŁo)? Argument readings indicate +R.

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

Occupations are object; offices are role. Padeiro, médico, advogado, motorista, professor pass the bare predicative (ela é médica) and require no particular second entity: practising a trade is participation, not dependence. An office or rank held in an institution is role: presidente, senador, vereador, cardeal, abade, coronel, delegado. Each takes an argument de-genitive (presidente da empresa, abade do mosteiro), and the office cannot exist without the institution. The same goes for agentive names tied to one event or relatum (comprador, autor, vítima, acusado, paciente, sócio), whether or not they also name a job. The Pessoas_por_vocação frame does not decide the type; each noun takes this test.

Parts are object; fractions are role. A body part or an artifact part (dedo, tronco, gengiva, roda, telhado) has an identity of its own and is object, although no finger exists without a hand. The dependence test targets concepts defined by a relation to another entity, and being a part of something is not such a relation. A noun that names a fraction of whatever whole is given (metade, pedaço, fragmento, trecho, décimo) is role (§6.3).

Sports are event. A sport or discipline is a kind of activity (atletismo, maratona, salto em altura, squash), and so is a competition (torneio, campeonato, oitavas). Team positions (goleiro, centroavante, lĂ­bero) are role: each is defined in the team.

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, without changing what the concept is. Duas destruiçÔes, 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 diverges is a lemma with a product sense: a participant of the event — typically its Result or Product — with existence independent of the event, persisting 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 passed by the product sense and failed by the event sense:

  1. Spatial location. A pintura estå na parede. / *A destruição estå na parede.
  2. Physical or material predicates. A tradução tem 300 påginas. / *A destruição tem 300 påginas.
  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 — a frame-layer fact, not a type-layer one.

Grimshaw's complex-event/simple-event/result trichotomy is not used here: it separates nominals by argument structure, not by type (Appendix §A.4).

quality / value#

Ask whether the lemma names the dimension or a region within it (Appendix §A.1). 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.

The type does not encode causal origin. Whether a quality resulted from an event (cansaço, force-maintained postura) or not (beleza, inteligĂȘncia) plays no role in the classification: both are quality, and their values — cansado, quebrado, sentado on one side, bonito, inteligente on the other — are both value. That distinction belongs to the situation: plain attribution, active maintenance and resultant conditions must be distinguished from their definitions. Maintained and resultant readings can both belong to State; plain properties belong to the Attribute family. Sharing a lemma type does not establish that any specific allocation is correct. See Appendix §A.5 for the full argument, including why no reliable test separates a "resultant" quality from a non-resultant one.

Referential use does not create a new type. O bonito, um americano refer to an entity characterized by the value without the lemma becoming a different kind of thing — see §1.4, "Substantivized adjectives used elliptically," for the construal-shift diagnostic that tells this apart from genuine lexicalization (dependenteₙ/dependenteₐ, §7).

Health conditions are quality; episodes are event. A disease or condition (bradicardia, catarata, alcoolismo, glossite, diabetes) is a categorial dimension that inheres in a bearer, typed like cansaço. Its values (doente, diabético, enfermo) are value. A noun naming an occurrence of the condition (surto, crise, ataque, piora, recaída) is event, like any happening. A lesion, mass or body substance is a thing, not a condition: cisto, cålculo, pólipo, variz, pus, catarro are object, because they pass the spatial-location test (o cisto estå no ovårio).

Existing NOUN lemmas for value categories stay value. Demonyms and nouns for adherents, ages, ethnicities and orientations (brasileiro, judeu, católico, idoso, gay) can exist as NOUN lemmas that duplicate an ADJ value lemma. They keep the value type, as the construal-shift analysis gives the adjective (§1.4, §7). Merging their LUs onto the ADJ lemma is LU work, not a typing decision. The doctrine or group itself (budismo, catolicismo, judaísmo) is object, like lei or ideia. Nouns with no value reading (criança, recém-nascido, mendigo) are sortals, typed by the object/role test.

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; the link between the two is dul:isDefinedIn (§3), 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 construal shift, not a second type. The lemma remains role (§1.4).

5. Procedural types#

Nine types. Procedural lemmas take no DUL reference.

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 — individuals, times or occasions todo, cada, algum, vĂĄrios, ambos, tudo, qualquer, alguĂ©m, algo, sempre, geralmente, Ă s vezes
polarity assert or deny instantiation sim, nĂŁo, nunca, jamais, nada, nenhum, nem, sem
degree how to locate a value relative to a standard muito, pouco, mais, menos, bastante, demais, tĂŁo
modality modal status of the proposition — epistemic, deontic, ability or bouletic (wish) — with no scale named talvez, necessariamente; modal auxiliaries (poder, dever, ter que, precisar); optatives (oxalá, tomara, Deus queira, quem me dera); free mood markers
connection inferential or discourse relation between units contudo, portanto, porém, entretanto, pois, se
focus information-structural instruction, including verum focus and exception só, também, até, mesmo, sequer, inclusive, exceto, salvo, de fato, realmente
predication how to build the predication; contributes no concept beyond what the head, frame or construction supplies copulas (ser, estar), tense and aspect auxiliaries (estar + gerund, ir + infinitive, ter + participle), support verbs, case-marking de/a/por
interaction how to manage the exchange — appeal, stance display, information source or perspective, hedging, comment on the speech act nĂ©, hein, viu, tĂĄ, uĂ©, nossa, olha, tipo, assim, diz que, supostamente, segundo, francamente

The nine types are organized by functional category; their theoretical provenance and relation to Relevance Theory is in Appendix §A.6.

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 locate a region against a contextually supplied standard → degree.

Indefinite pronouns are quantifiers. Tudo, tudo que, qualquer, qualquer um, qualquer coisa, a qualquer momento, våria, algo and alguém quantify over a domain of entities or times, exactly as the determiners todo and algum do (algo is alguma coisa, alguém is alguma pessoa). They are quantification in every POS. The negative ones (ninguém, nada, nenhum) deny existence and are polarity (§6.1). Personal, demonstrative, relative and interrogative pronouns stay reference, and so does the indefinite article um (§6.3).

The same holds for indefinite place adverbs. Algures, alhures, dalgures and dalhures ("somewhere", "elsewhere") quantify existentially over places, as algo does over entities. They are quantification, not reference: nothing is resolved against the utterance situation, as it is for aqui or lĂĄ. The negative nenhures is polarity.

polarity vs focus: who carries the negation? A negative scalar item is polarity only if it carries the negation itself. Nem sequer, nem mesmo, nem ao menos and tampouco (também não) do, so they are polarity. Sequer alone needs a negation found elsewhere (não conseguiu [sequer] apresentar um plano), and adds only scalar "even" over alternatives, like até and mesmo. It is focus.

Nem carries the negation itself in all its uses: as a negative coordinator (nĂŁo come [nem] bebe, [nem] A [nem] B), as a standalone negator ([nem] pensar) and as negative "even" ([nem] ele sabe). It is polarity in every POS, and so is e nem. Coordination is the construction it occurs in, not what the lemma encodes.

Exception and inclusion are focus. Exceto, salvo, tirante, excluindo, com exceção de and incluindo remove members from, or add members to, a set of alternatives, as exclusive só and additive inclusive do. They are focus in every POS. The conditional salvo se and exceto se ("unless") are connection.

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 (§6.2).

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 (provåvel, certo), and stand to it exactly as belamente stands to beleza (§6.1). The following diagnostics help 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 nĂŁo Ă© provĂĄvel que
 ✓ *nĂŁo talvez
 ✗
governs mood no yes — subjunctive

Degree modification supports a value analysis when it modifies a lexicalized scale. It is not decisive alone: quotation, coercion and contextual readings can alter acceptability. Talvez normally lacks this scalar behavior and can license the subjunctive. Modal auxiliaries remain modality on their modal readings. Use the diagnostics together with lexical meaning.

The tests apply to the head of a multiword expression. The certainty MWEs com certeza, de certeza, por certo, ao certo and decerto contain the dimension itself (certeza, certo). They are gradable through their head (com toda certeza), even though *muito com certeza is blocked by syntax. Like certamente, they name the top of the certainty dimension, so they are value. The necessity adverbs necessariamente and obrigatoriamente pass none of the tests (*muito necessariamente): they express a modal force with no scale named, so they are modality.

The modal flavour is a fact about the occurrence, not the lemma. One modal auxiliary is one modality lemma, whichever flavour it expresses: epistemic ([deve] ter sido no meio do ano), deontic ([deve] ser administrada por via oral), or ability and circumstantial possibility (não [podendo] el-rei alcançar, quem se deslocar a pé [pode] chegar). Following Kratzer, a modal contributes a quantificational force (dever necessity, poder possibility), and the context supplies the kind of modality. Splitting by flavour would fail §1.4's own criterion: the flavours can be readings of one typed lemma rather than different lemma types. They still require distinct LU/frame analyses where their meanings differ; mixed deontic and epistemic annotations are not validated by a shared modality type. The flavour is recorded in the frame the LU evokes: Capability, Possibility, Required_event, Being_obligated, Likelihood, Expectation. A modality lemma may carry these LUs like any other lemma (§8).

What splits off is a reading that is not modal at all, under §1.4: dever-se a "be due to" and dever "owe" require conceptual analysis of causation/dependency or indebtedness; do not assign event merely because the forms are verbs or occur in Commerce-related frames; the nouns o poder (power: Objective_influence, Electricity, the Poderes of the state) and o dever (duty, deveres) are separate NOUN lemmas.

Optatives are modality. OxalĂĄ, tomara, Deus queira and quem me dera express the speaker's wish about the proposition: bouletic modality, with no scale named. Like talvez, they govern the subjunctive ([tomara] que chova). They are modality in every POS, INTJ included. They do not manage the exchange, so they are not interaction.

The copula and support verbs are predication. They instruct how to assemble a predication and contribute no concept of their own. This does not prevent them from having LUs (§6.2, "Not an LU target").

Aspect: phase verbs are event, aspect auxiliaries are predication. The line does not depend on whether the verb heads a periphrasis. It depends on whether the verb names a phase of an event.

  • Phase verbs name the start, continuation, end, interruption or resumption of an event, or its habituality: começar, iniciar, continuar, prosseguir, acabar, terminar, parar, deixar de, passar a, voltar a, costumar. These phase/habitual readings are typed event, whether realized with nominal or verbal complements; this does not classify every other reading of the same forms. They evoke the phase frames (Activity_start, Activity_ongoing, Activity_finish, Activity_stop, Process_resume, Custom, 
) exactly as iniciar or interromper do. Começou a reuniĂŁo and começou a chover are one event lemma, and two LUs only if the frames differ.
  • Aspect and tense auxiliaries mark the event without naming a phase: estar / andar / vir + gerund (progressive; vem melhorando), ir + infinitive (future), ter / haver + participle (perfect), estar prestes a (prospective). They are predication. An LU annotated on such an auxiliary — the progressive estar in Activity_ongoing — stays on the predication lemma.

Two supporting diagnostics for the phase-versus-auxiliary distinction:

Test phase verb — começar auxiliary — progressive estar
takes a nominal complement with the same meaning começou a reunião ✓ estar + NP has no progressive reading ✗
a lexical verb with the same phase meaning can replace it começou a reunião ≈ iniciou a reunião ✓ no lexical verb means "in progress" ✗

A verb with both uses splits under §1.4 only if both readings carry LUs. Ficar in "[ficou] trĂȘs dias internado" or "nada [fica] na mesma" means remain (State_continue, event). In "[ficava] voando" it means keep doing, a phase (Activity_ongoing, also event). In "ficou rico" it is the inchoative event reading. The predication ficar keeps only its copular-stative LUs.

interaction is about the exchange, not the proposition. It covers interjections (uĂ©, nossa, ai), interactional and tag particles (nĂ©, hein, viu, tĂĄ), attention markers (olha, escuta), hedges and approximators (tipo, assim), and quotative or evidential markers (diz que, dizem que) — whatever manages the interaction rather than contributing to or operating on the proposition expressed.

Three boundaries:

  • vs connection. A connective relates two propositions inferentially (portanto, contudo); an interactional particle relates the utterance to the hearer (nĂ© 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, supostamente, alegadamente, pretensamente, aparentemente, presumivelmente). FNBr keeps evidentials in interaction because they invoke the hearer's assessment of the source rather than stating the speaker's confidence (Appendix §A.7). That holds for reportative and inferential evidentials alike, as clitic, verb phrase or adverb (parece que, pelo visto, ao que parece). Attribution of source or perspective by a preposition works the same way: segundo ("according to") and para in [para] a Dinamarca, era bom
 ("in X's view") are interaction. De acordo com, conforme and consoante split under §1.4. Their source reading ([de acordo com] a polĂ­cia) is interaction. Their conformity reading ([de acordo com] a lei, "in accordance with") names a relation of correspondence and is relation. Clausal conforme ([conforme] ele disse, [conforme] o tempo passa) stays connection.
  • vs focus. SĂł, atĂ©, mesmo operate on the proposition's information structure and are focus. Tipo and assim as hedges signal approximation in the exchange rather than restructuring the proposition.

Three more cases sit on these boundaries:

  • The verb parecer. Copular parecer (parece cansado) is not a bare copula: it adds the appearance the predication rests on. It stays event. The evidential parece que is interaction.
  • Speech-act adverbs (francamente, sinceramente, honestamente). In sentence-initial use ([Francamente], nĂŁo sei) they comment on the speech act, not on the proposition, so they are interaction. Their manner reading (falou francamente) is a value, like any manner adverb. When both readings carry LUs, the lemma splits under §1.4.
  • Confirmation adverbs. De fato, realmente, efetivamente, verdadeiramente and deveras mark verum focus: they put the truth of the proposition in focus (eles, [de fato], sĂŁo o povo mais feliz), so they are focus. Na verdade and na realidade introduce a correction or reformulation of a previous unit (nĂŁo Ă© X; na verdade, Ă© Y), so they are connection. Before an adjective, realmente and deveras are intensifiers ([deveras] abstratas), which is degree. That reading splits off under §1.4 when it carries LUs.

interaction lemmas evoke pragmatic frames like any other lemma.

6. 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; ADV, ADP and NUM need worked guidance.

Three POS sit at the edge of the type layer:

  • PROPN lemmas are typed by what they name. Institutions, places, peoples, people and brands are object (UFJF, Ibitipoca, Cemig); an event is event (a festival). An acronym of a common noun is not a proper name: PIB, CPF and PEC are NOUN, typed like the noun they abbreviate.
  • PUNCT and SYM lemmas carry no type. They encode neither a concept nor a processing instruction about the proposition, so the two-sort vocabulary of §2 does not apply.
  • Digits, times, percentages and roman numerals (18, 18h, 15%, XVIII) are NUM lemmas and value, like the numerals of §6.3. Written ordinals (1Âș, 15ÂȘ) are ADJ and value.

6.1 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. 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, ocasionalmente, regularmente) a region on the frequency dimension value
Degree / intensity (muito, pouco, bastante) locate a value against a standard degree
Universal frequency (sempre, diariamente, o tempo todo) universal quantification over times or occasions quantification
Generic frequency (geralmente, normalmente, usualmente, habitualmente, em geral) generic quantification over occasions quantification
Existential frequency (Ă s vezes, por vezes, de vez em quando, volta e meia) existential quantification over occasions quantification
Phasal (ainda, jĂĄ) quantification over the phases of a state, relative to a reference time quantification
Epistemic, scalar (provavelmente, certamente, possivelmente, com certeza, obviamente) a region on the probability/certainty dimension value
Epistemic, non-scalar (talvez) epistemic status, no scale named; governs subjunctive modality
Necessity (necessariamente, obrigatoriamente) modal force, no scale named modality
Evidential (supostamente, alegadamente, aparentemente, pelo visto, ao que parece) source of the information interaction
Speech-act (francamente, sinceramente, honestamente), sentence-initial comment on the speech act interaction
Confirmation (de fato, realmente, efetivamente) verum focus focus
Reformulation (na verdade, na realidade) correction of a previous unit connection
Epistemic stance as a concept (dĂșvida, incerteza, certeza) the dimension itself quality
Place and deictic time (aqui, hoje, agora) resolution instruction reference
Time of day, anchored time (de manhĂŁ, Ă  noite, de noite, mais tarde, com o tempo) temporal location of the event, nothing resolved against the utterance relation
Conjunctive (contudo, portanto) discourse relation connection
Affirmation / negation (sim, nĂŁo, 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 encode an instruction to resolve a referent against the utterance situation — the textbook case of procedural encoding. Their image-schematic character is carried by what their frames ground in (FIGURE_GROUND, DEICTIC).

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. Nada is polarity.

Frequency adverbs require both scope and degree diagnostics. Their types cannot be read from a shared frame name. Use §5's diagnostics alongside the quantificational or scalar meaning; a Frequency allocation is not automatic approval of every candidate reading.

  • Gradable (muito frequentemente, muito raramente, muito ocasionalmente, muito regularmente ✓): these name a region on the frequency dimension, the counterpart of frequente, raro or regular, and are value.
  • Not gradable (*muito sempre, *muito geralmente, *muito Ă s vezes, *muito diariamente ✗): these quantify over occasions, in the sense of Lewis's adverbs of quantification, and are quantification. The force is universal (sempre, diariamente, o tempo todo), generic (geralmente, normalmente, usualmente, habitualmente, tipicamente, em geral) or existential (Ă s vezes, por vezes, de vez em quando, volta e meia).
  • Zero frequency (nunca) stays polarity, by the rule above.

The type follows the frequency reading only. Normalmente also has a manner reading (as pessoas trabalham normalmente, "in the normal way"), which is value. When both readings carry LUs, §1.4 splits the lemma.

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).

6.2 Prepositions and adpositions#

ADP is one UPOS covering lemmas on both sides of the conceptual/procedural line; 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, apĂłs, dentro de, ao lado de, antes de relation
Topic sobre, a respeito de, acerca de, quanto a, em relação a, com relação a ("um livro sobre linguística", "sobre as fake news, 
") relation
Means or route por ("por via oral", "pela estrada"), em ("fala em inglĂȘs"), atravĂ©s de relation
Having or wearing com ("mulher com blusa rosa") relation
Privative sem polarity
Causal, concessive, conditional por causa de, apesar de, devido a, em vez de, causal por (preso por roubo) connection
Conformity de acordo com, conforme, consoante ("de acordo com a lei") relation
Exception, inclusion exceto, salvo, tirante, excluindo, incluindo, com exceção de focus
Source or perspective segundo, de acordo com, conforme, consoante ("de acordo com a polĂ­cia"), para ("para mim", "para a Dinamarca") interaction
Marking an FE that a head selects de in gostar de, a in obedecer a, em in acreditar em, com in contar com, de in orgulhoso de predication
Marking an FE that the head's frame supplies por (passive agent), com (instrument), para (beneficiary), de in pai de JoĂŁo and o retrato de JoĂŁo predication
Anchoring an N de N construction de with a sortal head: mesa de madeira, o carro de JoĂŁo, measure um copo de ĂĄgua predication

The last three rows are the case-marking uses. Portuguese nouns have no case morphology. In Latin or Finnish, case affixes signal how a noun phrase relates to the word it depends on. In Portuguese, prepositions do that job.

What a case-marking preposition contributes. Three claims about such a preposition must be kept apart:

Claim Holds?
(a) the use is not an LU target yes: an LU would duplicate the head's valence
(b) the preposition evokes no frame of its own yes: the frame belongs to the head or to a construction
(c) the preposition contributes no meaning no

Neither (a) nor (b) entails (c). Doing the job of case does not make a form empty: Jakobson's analysis of the Russian cases (1936) gives every case a general meaning. FNBr takes a schematic view. De has one underspecified meaning, roughly "intrinsic association or dependency", historically derived from source → partitive → genitive (compare Langacker 1992 on English of). The specific relation (possession, material, measure, authorship) is not stored in de. It comes from the head's frame, from a construction, or from pragmatic enrichment.

Two alternatives are rejected. Vacuity (de means nothing) is contradicted by the data below. Polysemy (one sense, and one LU, per relation) would copy the FE inventory of every head frame into the preposition, although the readings are predictable from head and complement.

The evidence. Vary one side at a time.

  • Hold de fixed and vary the nouns. Mesa de madeira (material), mesa de JoĂŁo (possession), mesa de jantar (purpose), mesa de dois metros (measure): the relation tracks the nouns, so it is not stored in de. O retrato de JoĂŁo is ambiguous between JoĂŁo as painter, as the person depicted and as owner. De is compatible with all three, and each is an FE of the head's frame.
  • Hold the nouns fixed and vary the preposition. Mesa de madeira ≈ mesa em madeira shows no truth-conditional difference. But copo de ĂĄgua allows the measure reading ("a glassful"), while copo com ĂĄgua forces container plus contents. De is doing something there, even if part of the measure reading belongs to the polysemy of copo.

Four kinds of prepositional contribution.

Kind Example Who evokes the frame Where the preposition is recorded Type
A. Frame-evoking sobre/sob a mesa, durante, apĂłs, topic sobre, means por, mulher com blusa the preposition its own LU, with the frame's FEs relation
B. Head-selected gostar de, acreditar em, contar com, orgulhoso de the head the head's valence: phrase type PP[de] predication
C. Frame-supplied passive por, instrument com, beneficiary para; pai de JoĂŁo, o retrato de JoĂŁo the head, whose frame has the FE the head's valence; the FE assignment resolves any ambiguity predication
D. Constructional anchor mesa de madeira, o carro de JoĂŁo, measure um copo de ĂĄgua an N de N construction the construction, as its construction-evoking element predication
  • A. The preposition names the relation. Replacing sobre with sob changes what is said, so this use is not case-marking. Topic sobre, means por and the com of having belong here too. The PP answers a question with the preposition inside it (Sobre o quĂȘ?, Por onde?, Com que roupa?), and it can stand with no head that selects it (Sobre as fake news, responda
).
  • B. The choice of preposition is a lexical fact about the head. Using another one gives an ungrammatical string, not a different meaning (*gostar em mĂșsica). What the position means (what is liked, who is obeyed) is an FE of the head's frame. The choice is often motivated (acreditar em and pensar em construe the content as a location), but that motivation is systemic, not something a single token predicts.
  • C. No single head selects the preposition: any passive verb takes por. The role the PP fills already belongs to the head's frame, and the preposition picks it out or allows it.
  • D. The head noun's frame has no FE for the relation, so the relation comes from a construction. FrameNet's constructicon (Fillmore, Lee-Goldman & Rhomieux 2012) lets a construction evoke a frame, with a lexically fixed construction-evoking element. In FNBr, frames evoked by constructions are recorded with rel_evokes. In N de N, de is that element: it carries meaning as the fixed anchor of a meaningful construction, without being an LU.

Telling C from D. Does the head noun's frame already have an FE that the de-phrase fills?

  • Yes → C. Retrato has Creator and Depicted; pai has its relatum; a deverbal noun has its event's FEs.
  • No → D. The relation must come from a construction.

Partee & Borschev's analysis of the genitive draws the same line. With a relational head (pai de João), the head supplies the relation and the genitive marks its argument. With a sortal head (o carro de João), the construction introduces a free relation R, resolved by context or by the head's qualia (Vikner & Jensen 2002). Adding peripheral FEs such as Material or Possessor to every artifact frame would turn D into C, but it repeats the same FEs in every sortal frame. The construction states the generalization once. Which option FNBr adopts is open (§A.9).

Diagnostics for kind B. The three diagnostics below separate kind A from kind B: a preposition that names a relation from one that a head selects. They do not apply to the nominal uses. De quem Ă© o carro? — De JoĂŁo passes questionability. Copo de/com ĂĄgua passes contrastive substitution while mesa de/em madeira fails it. A sortal head like mesa lists no de in its valence, so government does not apply. For nominal uses, use the C/D test above.

  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 — 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, predication regardless of what it means elsewhere.

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

Not an LU target, which is not the same as contributing nothing. Three points follow from the four kinds.

  1. A case-marking use is not an LU. In gostar de mĂșsica, the frame is evoked by gostar. De mĂșsica is annotated as an FE of that frame, with de as part of its phrase type (PP[de]). The preposition is recorded in the head's valence pattern as the marker of that position, the role synsem:marker plays in OntoLex-Lemon. An LU for de in gostar's frame, or in some "genitive" frame, would record the same valence fact twice and treat a schematic contribution as a frame of its own.
  2. In kind D, the frame belongs to the construction. De anchors the N de N construction that evokes it, and is not an LU of that frame.
  3. The predication type does not forbid LUs. The type says the lemma contributes no concept of its own that the head, the frame or a construction does not already supply. It does not say the lemma is meaningless, nor that it can never be a frame target. §8 holds for this type as for every other. Some readings of predication lemmas are annotated as targets and carry LUs: the progressive estar in Activity_ongoing, and copular-stative ficar and estar in State frames (§5).

So "not an LU" describes a use (kinds B, C and D), not a type. Copulas and support verbs work the same way. In their ordinary use they assemble a predication around a noun or adjective that evokes the frame. They have LUs only where a reading of their own is annotated as a target.

An existing LU on a predication lemma is therefore something to review case by case, not an error in itself. If the LU annotates a kind B or C use (the FE's preposition annotated as if it were the target), it duplicates valence. If it annotates a kind D use, the frame belongs to the construction. If it annotates a reading that is itself a frame target, it stands.

Outcome of the 2026 review. The LUs on predication prepositions were reviewed against these rules. Em in Using and Separating and instrument com in Using repeated the head's valence and were retired. Com in Wearing, por and em in Means and the topic LUs name a relation of their own and moved to relation lemmas. De in Partitive, Possession and Type moved to the complex lemma $n_de_n, which stands in for the N de N construction (§A.9, item 14). Partitive sets whose head noun evokes a frame of its own (início, metade, maioria, the body parts) went to the head's LU or were dropped as duplicates of it.

Annotating N de N. A complex lemma with an LU gets a span on the annotation page under the words it covers (ADR 0008). Clicking the span offers its LUs. The target is de alone, as in the sets moved in 2026: tick de and untick the span in the span bar. The two nouns carry the frame's FEs.

The pattern of $n_de_n is

({NOUN} | {PROPN} | {NUM} | {PRON}) [{ADV}] [{ADJ}] [{ADJ}] {=de} [{$quant}] [({$art} | {$dem} | {$poss_det})] [{$num}] [{ADV}] [{ADJ}] ({NOUN} | {PROPN})

{=de} also matches the contractions (do, da, dos, das), which are forms of de. The head side allows um dos, algumas das and post-nominal adjectives (pontos fortes do). Tested in November 2026 on the 128 sets of the three LUs, it covered 117 of them. It misses coordination (da PM e da prefeitura), a comma between the nouns, and an N de N with no head noun (dos 36 clubes, 11
). It also matches names such as Copa do Mundo and Rio Grande do Norte, which have lemmas of their own.

What the preposition itself construes in kinds B and C is not recorded today. A per-annotation label for it, alongside the FE, is an open proposal (§A.9).

Applying the rule. Adpositions illustrate §1.4's one-type-per-lemma principle. Split only when type divergence is established within the lexical identity. Splitting stays bounded because readings that share a type do not split: de has at least six distinguishable readings but only two types in the analyses below. All the case-marking uses of a preposition (kinds B, C and D) are uses of its one predication lemma. It is the same lemma with the same schematic contribution in every case; the specific relation is evoked by the head or by the N de N construction. None of these uses needs an LU. Splitting the nominal uses off as a relation lemma with an underspecified relation is not adopted: it would add splits (§A.9, item 7) for a generalization the construction already states.

Lemma Type Covers Examples
de₀ predication head-selected (B); genitive with a relational head (C); material, possession and measure with a sortal head (D) gostar de mĂșsica; o pai de JoĂŁo; mesa de madeira; o carro de JoĂŁo; um copo de ĂĄgua
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; Ă s trĂȘs horas
em₀ predication government acreditar em algo; pensar em vocĂȘ
em₁ relation containment, location, temporal location, means na mesa; em 2020; no Brasil; fala em inglĂȘs
por₀ predication passive agent; government; price marking destruído por João; optar por outro; comprei por dez reais
por₁ relation path, duration, means, stance passou por SĂŁo Paulo; por trĂȘs horas; por via oral; manifestação por melhorias
por₂ connection cause, linking an event to its cause preso por roubo; não veio por estar doente
com₀ predication government; instrument and manner marking contar com vocĂȘ; cortou com a faca; com cuidado
com₁ relation comitative, accompaniment, having or wearing foi com Maria; cafĂ© com leite; mulher com blusa rosa
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
para₃ interaction perspective ("in X's view") para a Dinamarca, era bom

sobre relation superposition and topic — no split, both readings have one type o livro está sobre a mesa; um livro sobre linguística
sem₀ polarity privative — no split, one type covers every reading sem açĂșcar; saiu sem pagar

Sobre and sem do not split. Sobre has two readings, superposition and topic, each with its own LU, but both are relation. The split is driven by type divergence and nothing else.

Contractions are forms, not lemmas. Do, ao, pelo, no, num, dum, pro and coa fuse a preposition with an article. UD splits the token into its two words, and the article is a form of o or um (reference). The preposition part is a form of the preposition's lemmas: do/da/dos/das belong to de₀ and de₁, ao/à/aos/às to a₀ and a₁, pelo/pela to por₀–por₂, no/na/num/numa to em₀ and em₁. Contractions get no lemmas of their own, so the split covers them without being repeated.

Disambiguation. Splitting means UPOS filters nothing between de₀ and de₁ — they share both text and POS. Disambiguation comes from the syntactic context, using evidence already in the UD parse:

  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 — this supports the predication reading. Parser and valence evidence remain defeasible; preserve alternatives where government is uncertain.
  2. Complement semantics. A place or time complement favours the relation lemma; an infinitival complement under para forces para₂.
  3. Unresolved. Where the evidence does not distinguish the candidates, preserve the candidate set and record uncertainty. Frequency or an assumed lower cost of deleting conceptual content is not semantic disambiguation.

Extended discussion of the accuracy trade-off this moves to the parser is in Appendix §A.8.

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: a small set of very frequent forms each span the conceptual/procedural line. The split count is concentrated — a handful of function words account for most of it.

Complex prepositions are MWE lemmas tagged ADP. Dentro de, ao lado de, por causa de, apesar de take entry_type = mwe and pos = ADP, an exception to the X convention for MWE entries (§1.2). A multiword expression that takes no complement (daqui em diante, em pleno ar) is an adverb, ADV. 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 — the tag predicts the type in neither direction.

Consequence for participant structure. Review participant associations where appropriate role lexicalizations exist. A spatial or temporal relation does not require a Portuguese noun for every FE. Missing lexicalization is not a type error, and role definition does not require the participant LU to evoke the relation frame.

6.3 Numerals and proportional nouns#

NUM looks quantificational, since dois livros behaves like alguns livros, but the type follows what the lemma names.

Class Examples Type
Cardinal numeral um, dois, cem, mil, milhĂŁo (NUM) value
Ordinal numeral primeiro, quarto, décimo value
Collective numeral noun dezena, dĂșzia, centena, milhar, milhĂŁo, bilhĂŁo (NOUN) value
Dual quantifier ambos (every POS) quantification
Proportional noun maioria, minoria, metade, parte, totalidade, resto, porção role

Numerals name a point. A cardinal names a point on the number dimension and an ordinal a point on an ordering. Both are nameable without a bearer (dois Ă© par, o primeiro chegou), and both evoke Cardinal_numbers or Ordinal_numbers as concepts. They are value, like any region or point. The determiner-like use (dois livros) is a construal of that value by the Num + N construction, not a second type. The collective nouns name the same points (uma dezena is 10) and are value too.

ambos is a quantifier, not a number. It has no number-name use (*o ambos, *ambos Ă© par). It quantifies universally over a domain of two, so it is quantification in every POS, NUM included.

Zero. Zero as a numeral names the point 0 and stays value. "Zero is polarity" (§6.1) concerns nada and nenhum, which deny existence; zero measures it.

Units and time intervals are value. A unit of measure names a fixed magnitude on a dimension: metro, litro, quilo, caloria, polegada, currency units (dĂłlar, real, centavo) and measure nouns (pitada, colherada, punhado). Dois quilos is a numeral plus a value, as dois livros is a numeral plus an object. Calendar and clock nouns name a stretch of the time line (dul:TimeInterval ⊑ dul:Region): janeiro, segunda-feira, dia, semana, manhĂŁ, verĂŁo, sĂ©culo, prazo. So do dates and holidays (feriado, Natal, aniversĂĄrio). A noun that names the celebration itself, as carnaval does, is event.

Proportional nouns are role. A half, a majority or a remainder cannot exist without a whole that is not part of it, and that is §3's external dependence test. Metade stands to its whole as filho stands to its parent. Their quantifier-like use (a maioria dos alunos, metade do bolo) is the partitive construction, recorded at the frame and construction layers, not by the lemma type.

7. Why coarse: the three layers#

The asymmetry is the most important property of this design; with the type on the lemma it reads as three layers:

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
  • An event lemma can have causative and inchoative readings in different frames without requiring a type split. A proposed resulting-condition reading still needs semantic evidence; the lemma's type neither validates nor rules out a State allocation (§8).
  • Aquecer is one lemma typed event, with LUs in Causar_mudança_de_temperatura and Mudança_de_temperatura. Same type, different frames — the ordinary case.
  • value behaves the same way across @attribute, @state, @entity, @experiential and (as adjectives and adverbs) @relation. 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). frm_people_by_origin is an @entity frame built over value lemmas, and it is not an exception — the same analysis can be evaluated for other participant-category frames. A frame's name or current type distribution does not establish this analysis for all its LUs. For a supported construal shift (§1.4), the lemma stays value and the referent varies with what the construction supplies.
  • POS differences split the lemma, not the LU. Dependenteₙ ("meu dependente") is role; dependenteₐ ("ele Ă© dependente") is value — two lemmas by §1.2, each with its own type and its own LUs. Unlike americano above, this is not a construal shift: the noun names a fixed institutional role, not "person" as a default referent for the value (§1.4).
  • 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.

Construal shifts change neither. Ele Ă© pai shifts a functional concept to a sortal predicate without changing the underlying concept (Löbner); o bonito and um americano refer to an entity a value lemma characterizes without becoming sortals themselves. Neither automatically establishes type ambiguity or requires a lemma split. Whether a distinct LU is needed depends on the established reading and frame, not on the presence of an NP construction alone. §1.4 has the general statement and the diagnostic that tells a construal shift apart from a genuine split like dependenteₙ/dependenteₐ above.

8. Lemma type and LU allocation#

The lemma type does not restrict which frames a lemma's LUs may evoke. A frame describes a situation, and one situation can be evoked by lemmas of diverse types: an event verb naming the happening, a role noun naming a participant, a value adjective naming a condition, a quality or relation noun naming a dimension or tie. No namespace × type compatibility matrix ("admit-set") is defined, and no combination of lemma type and frame namespace is an error by itself.

The namespace × type matrix proposed in Phase 2 of the structural repair (2026-10-14) is withdrawn as allocation policy. Its outcomes (PASS, CONDITIONAL, FAIL, EXEMPT, NO_RULE) carry no documentation authority and must not be used as review verdicts.

Type and namespace answer different questions. The type says what kind of concept, or processing instruction, the lemma encodes (§3, §5); the namespace says what kind of situation the frame describes (Frames). An LU's allocation is judged by its reading against the frame's definition and participant structure, not by its lemma's type.

Specific evaluations#

Particular combinations may later be checked by targeted evaluations, each with its own explicitly stated rule and rationale. The expected first case is FNBr's agentive-name policy: agentive nouns (trabalhador, comprador) with LUs in eventive frames are candidates for moving to Entity frames, with their participation recorded as an association (see below). A targeted evaluation produces a review list, not automatic moves, retyping or merges.

The second case, decided 2026-11-11, is the value adjectives in eventive frames. An adjective stays in the event frame when its reading evokes the event with its participants, so that the noun it modifies fills an FE of the event (arma letal, doença curåvel). A relational adjective that sorts the noun by domain (material escolar, exame médico) is Attribute.Classifying, in Domain. See Property adjectives and event frames.

What the type layer checks on its own#

Independently of frames, a lemma's type can be checked against the criteria of this document:

  • POS consistency: lexical verbs are event, except copulas, auxiliaries and support verbs (predication) and modal auxiliaries (modality) (§5).
  • Deverbal nouns are event unless they have a product sense (§4), so a deverbal noun typed quality or object (crescimento, asfixia, batismo) is a typing candidate.
  • Externally dependent nouns are role (§4), so relational nouns typed object (avĂŽ, metade) are typing candidates.
  • Dimensions are quality and regions are value (§4).
  • Untyped lemmas require a type; punctuation and symbols take none (§6).

These are checks on the type assignment, not on LU allocation.

Participant structure#

Where a relation or event has suitable participant lexicalizations, review their association with the relevant FEs. Role definition (isDefinedIn) and a participant LU's lexicalization home are separate facts. Roles outside Relation frames are not automatically errors. FNBr places agentive names in Entity frames and associates them with the event FE separately.

Use FE–Frame associations only if all relevant LUs of the Entity frame qualify; otherwise use LU–FE associations for the subset. Resolve nested CON expressions before reporting a missing association. Absence of a role noun for a spatial or temporal FE does not invalidate that relation. See FrameNet patterns.

Review outcomes#

Multiple namespaces under a lemma are normal when distinct readings justify them; namespace diversity alone is not a split criterion.

  • A wrong sense is not detected by the lemma's type: ser.v in Artes_cĂȘnicas, receber.v in a sports-move frame and torcer.v in Dar_forma need semantic review of the reading, whatever the type.
  • Within-class distinctions such as Agentive versus Change require semantic review.
  • Missing glosses, duplicate-looking records and attached annotations require separate evidence review; none establishes correctness or equivalence alone.
  • No check automatically assigns types, splits lemmas, moves LUs or merges records. Apply only approved, evidence-backed changes.

Appendix#

Secondary material: full definitions, theoretical provenance, extended diagnostics, design notes, and open questions. None of it is required to classify a lemma, but §A.1 underlies the criteria in §3.

A.1 Definitions#

Six notions from formal ontology and lexical semantics, used by the diagnostics throughout this document and by the criteria in §3.

Rigidity (Guarino & Welty, OntoClean). A property is rigid if 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), and semi-rigid if essential for some instances and accidental for others. Anti-rigid properties cannot subsume rigid ones — why pessoa is not a kind of agente. Rigidity is defined here but is not a typing axis (§A.2).

Identity and sortality. A property supplies identity if it carries a criterion for saying "the same one again"; such properties are sortals, answering what something is. Pessoa supplies identity; comprador inherits it from pessoa; vermelho has none. Every object instantiates exactly one identity-supplying kind. Both object and role are sortals (§3); what separates them is external dependence, not sortality.

Unity. A property carries unity if all its instances are wholes under one unifying relation (functional, topological, morphological, intentional). Pessoa carries functional unity; água is anti-unity — an arbitrary sub-portion of water is still water. FNBr does not type on unity; it explains why mass nouns need no separate type (§A.3).

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. This is the diagnostic that separates object from role (§3, §4).

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 (irmĂŁ, 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 (§4) where the ontological test is hard to apply.

Dimension and region (GĂ€rdenfors, 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 — the distinction underlying quality vs value (§3, §4).

A.2 Why rigidity is not a typing axis#

Rigidity (§A.1) is not used to type anything. Anti-rigid intrinsic sortals — criança, adulto, novato, defunto, concepts an individual can stop falling under while continuing to exist, with no second entity required — are typed object, the same as rigid sortals like pessoa.

The reason is annotator reliability. The dependence test (§4) asks a single existence question: is there some entity, not a part of this one, that must exist? Rigidity asks a modal question: could this individual cease to fall under this concept and still exist? Modal judgements of that kind can be unreliable for annotation, so rigidity is not a typing axis. Causal origin is likewise not a separate lemma-type axis (§A.5).

Consistent with DUL. dul:Concept ⊑ dul:SocialObject ⊑ dul:Object, so an anti-rigid intrinsic sortal falls under dul:Object whether or not it is named separately — a loss of specificity, not an error. It also repairs role: under a single dependence test, filho is unproblematic even though being someone's son is arguably rigid (you do not stop being your father's son when he dies) yet plainly externally dependent.

What is lost. Criança and pessoa carry the same type despite differing in rigidity, and an export to a rigidity-aware ontology (UFO, gUFO) would have to decide Kind vs Phase at export time rather than reading it off the stored type. If that export becomes a requirement, a dedicated phase refinement would require explicit approval for anti-rigid intrinsic sortals, preserving the coarse meaning of object assignments — a candidate list could be generated by querying object lemmas whose frames are @entity frames built over life-stage, career-stage or status dimensions.

A.3 No abstract type#

Abstract nouns need no type of their own. DUL's Abstract is not "abstract nouns" — it is value spaces: Region, TimeInterval, SpaceRegion, the space of natural numbers. dul:Object already covers social and cognitive entities (SocialObject, InformationObject, Concept, Description), so lei, ideia, teoria, poema are Objects in DUL already.

Fact, proposition and modal nouns require reading-level review. Do not type a possibilidade as event merely because it concerns a possible situation; a probability/possibility dimension may instead be quality. Its frame and lemma type answer different questions.

Mass nouns need no separate type either: ågua is anti-unity but a rigid sortal, and unity is not a typing axis (§A.1).

Dot objects (livro = physical ‱ information) are deferred: if co-predication comes to require it, add a boolean refinement on object — physical vs social-informational — not a base type (see also §A.9, item 4).

A.4 Grimshaw's nominal diagnostics#

Grimshaw's 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 (§4). They describe nominal syntax, not this classification.

A.5 quality and value do not distinguish causal origin#

Three reasons back the claim in §4 that causal origin plays no role in the quality/value classification.

It duplicates the frame. Whether a producing event is on record is a fact about the situation. Plain properties, actively maintained conditions and resultant conditions are distinguished at the frame layer. That distinction does not require separate quality or value lemma types.

Type and situation have different granularity. Quebrado can remain value while its frame describes a resultant condition. This does not establish that every participial reading belongs to that condition frame.

There is no reliable test. A distinction between a "resultant" quality (cansaço, force-maintained postura) and a non-resultant one (beleza, inteligĂȘncia) as a lemma-type opposition would re-lexicalize the stage-level/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.

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

A.6 Theoretical grounding for the procedural types#

The conceptual/procedural split is Relevance Theory's (§2). The nine subtypes are not — no RT publication proposes this inventory.

RT subdivides procedural meaning along two axes not used here: the target of the constraint (explicature, higher-level explicature, or implicature — Blakemore 1987; Wilson & Sperber 1993), and the cognitive subsystem the item triggers (Wilson 2011, 2016). The inventory here is organized by functional category instead, applying 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 coarsen 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; CurcĂł 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 nĂŁo/not a logical entry, i.e. conceptual departure
degree degree semantics (Kennedy; and GĂ€rdenfors for the scale) outside RT
modality RT treats only mood and modal particles as procedural RT-aligned for mood/particles; local extension to auxiliaries (§5)

quantification and polarity. Classical RT (Sperber & Wilson 1986, ch. 2) treats logical words as concepts with logical entries — a claim about how an item enters inference. The criterion used here is 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.

degree is not an RT category. The "region located against a standard" criterion comes from degree semantics, and the dimension/region framing from GĂ€rdenfors (§A.1). It is procedural because it instructs rather than names — the operational test of §2.

modality. Wilson & Sperber (1993) and Ifantidou (1993, 2001) treat sentence adverbials such as certamente and provavelmente as conceptual; Papafragou (2000) treats modal verbs as conceptual. In RT only mood and modal particles are procedural. FNBr's scalar adverbs are value (§5, "modality vs value"), which keeps this row RT-aligned. One residue: 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 predication would misdescribe them as contentless. Papafragou's own analysis supports the single-lemma treatment of §5: one modal lexical entry per verb, with the epistemic, deontic or ability flavour supplied by the context, as in Kratzer.

What the inventory still lacks. Sentence mood and type (declarative, interrogative, imperative) has no type of its own — currently spread across reference (quem) and modality, with free-word mood markers landing in interaction. Whether mood deserves its own type depends on how much of it FNBr lexicalizes; mood realized by verb morphology or intonation is not a lemma and is out of scope. Tense and aspect are procedural on most RT accounts. Bound tense and aspect morphology is out of scope for the same reason. Free-word aspect is covered in §5: auxiliaries are predication, phase verbs are event, and the phasal particles ainda and jĂĄ are quantification, following Löbner's analysis of them as phase quantifiers.

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 consumes it.

A.7 modality vs interaction for evidentials#

Diz que is 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. FNBr classifies evidential readings as interaction across forms — diz que and the adverbs supostamente, alegadamente, pretensamente, aparentemente, presumivelmente (§5). The deciding point is that they direct the hearer to the source of the information, which the speaker may not vouch for, rather than grading the speaker's own confidence.

A.8 The cost of splitting adpositions#

Splitting moves work from the lexicon to the parser: the lexicon records a stable fact and the parser resolves a contextual one. That is the right direction, but it is a real cost and should be measured on the adposition tokens specifically, not averaged into overall lemma-resolution accuracy — see §A.9, item 9.

A.9 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 §4 give an external anchor; the number has not yet been measured.
  2. Rigidity is not recorded. See §A.2.
  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 open.
  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 (§1.5).
  6. The two-sort split has a boundary. Some lemmas plausibly encode both conceptually and procedurally (sempre, sĂł). 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 (§6.2). 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 — needs revisiting.
  8. Relatedness between split lemmas is unrecorded. Negação act/operator and pintura activity/canvas are 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.
  9. Adposition disambiguation accuracy. Splitting de, a, em, por, com and para by type (§6.2) moves their disambiguation to the parser, where UPOS cannot help because the candidates share it. Accuracy on adposition tokens should be measured separately, and the government heuristic evaluated on its own. If it underperforms badly, the alternatives are a richer valence-driven disambiguation or a coarser type inventory for adpositions — not a return to per-LU typing.
  10. The modality/value line needs validating, not deciding. The decision is taken (§5): scalar epistemic adverbs are value, talvez and the modal auxiliaries are modality, on 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 diagnostics over a closed list of epistemic adverbs before annotation opens.
  11. The modality/interaction boundary for evidentials. Evidential readings are interaction (§A.7); coverage still requires validation.
  12. interaction has not been tested at scale. It covers FNBr's pragmatic functions; its three boundaries — against connection, modality and focus (§5) — 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.
  13. The construal-shift diagnostic for substantivized adjectives is unvalidated. The "recoverable covert sortal" test (§1.4) separates productive ADJ→N ellipsis (o bonito, um americano) from lexicalized conversion (dependenteₙ, and presumably empregado, acusado and similar deverbal/deadjectival nouns). Coverage across the lexicon is unmeasured. The "people by X" @entity frames (§7) present a conventionalized middle case — a nationality, religion or vocation adjective defaults to a person referent systematically enough to be frame-licensed, yet no lemma split is made — and it is untested whether other systematic default-referent classes exist that the general test would miss.
  14. Where the N de N relation lives. For kind D uses of de (§6.2), the relation could be a frame evoked by an N de N construction, peripheral FEs (Material, Possessor) added to the head frames, or a free relation R resolved pragmatically. §6.2 prefers the construction, but this needs testing on corpus data against the coverage of FNBr's constructicon.
  15. Recording what a preposition construes. In kinds B and C (§6.2), the FE comes from the head, but the preposition still has a construal: acreditar em and pensar em treat the content as a location; depender de treats it as a source, where English depend on treats it as support. The SNACS scheme (Schneider et al. 2018; Hwang et al. 2017) gives every adposition token two labels: a scene role, which in FrameNet terms is the FE, and a function, the adposition's own construal. An optional coarse function label on the annotation would record this at the annotation layer without adding LUs. Spatial schematic meanings (SOURCE for de, CONTAINMENT for em) could instead be recorded as image-schematic grounding, as for deictics (§6.1). Neither is adopted yet.
  16. Means por and instrument com. §6.2 treats por via oral as kind A and cortou com a faca as kind C, yet many heads have a Means or Instrument FE that either PP could fill. The working difference is that means por also appears with nominals that do not select it (tentativa de suicídio por ingesta) and has a relation neighbour in através de. Whether that holds across the corpus, or whether instrument com should follow, is untested.