Nanoarguments: Vocabulary Survey and Model Specification

Schema.org

https://schema.org/

Summary: Schema.org is a collaborative vocabulary maintained by Google, Yahoo, Microsoft, and Yandex, primarily used for marking up structured data on web pages so search engines and other consumers can interpret content. It defines a hierarchy of types (with schema:Thing at the root) and a flat set of properties, deployed in JSON-LD, Microdata, and RDFa across a substantial fraction of the web.

Schema.org's coverage is broad (events, places, products, creative works, medical concepts, organizations, persons) and growing through a public extension process. Several of its types are directly relevant to scholarly discourse, including schema:CreativeWork, schema:Article, schema:Statement, and schema:Claim.

Core terms (relevant subset):

• schema:Statement: "a statement about something, for example a fun or interesting fact." Extends schema:CreativeWork.
• schema:Question: "A specific question - e.g. from a user seeking answers online, or collected in a Frequently Asked Questions (FAQ) document." Extends schema:Comment.
• schema:Claim: "a specific, factually-oriented claim that could be the itemReviewed in a ClaimReview." Targeted at fact-checking contexts.
• schema:CreativeWork: the broad parent class for authored works, with properties for author, date, license, and so on.

Limitations / scope notes:

schema:Statement is currently in schema.org's "new" area, meaning its definition may evolve as implementations and feedback shape it. Adopters should track changes.

The distinction between schema:Statement and schema:Claim is formality and intent: a Claim is the kind of factually-oriented assertion that gets reviewed by fact-checkers, while a Statement is broader and includes everyday observations, opinions, hypotheses, and other contributions that don't necessarily commit to being factually evaluated.

Schema.org's relations vocabulary is sparse for discourse purposes: it covers authorship, dates, licensing, and structural document properties, but not rhetorical relations between contributions. A model that uses schema.org for typing typically needs to combine it with another vocabulary (CiTO, PROV-O) for relations.

Web Annotation Data Model (W3C Rec, 2017)

https://www.w3.org/TR/annotation-model/

Summary: The Web Annotation Data Model is a W3C Recommendation defining a structured way to associate a body of information with a target resource. It supports arbitrary targets (documents, parts of documents via selectors, other annotations, arbitrary URIs), multiple bodies, and motivation typing through oa:Motivation and the oa:motivatedBy predicate. It is the foundation of W3C's annotation work, including the Web Annotation Protocol (delivery) and Web Annotation Vocabulary (extended terms).

Core terms:

• oa:Annotation
• oa:hasBody / oa:hasTarget
• oa:Motivation
• oa:motivatedBy
• oa:hasSelector (text-quote selector, XPath selector, etc.; see selectors)
• oa:purpose (role of a body vs motivation which is the role of the whole annotation - enables multi body annotations with distinct roles)

Limitations/gaps:

Web Annotation's scope is body-to-target anchoring with typed motivation, with optional selectors for parts of the target. This model uses oa:Annotation where this scope fits and uses direct triples between content nodes for rhetorical relations where no anchoring is involved (Relations on statements, Discourse and annotation layers).

oa:motivatedBy specifically expects an oa:Motivation instance as its object. Using it to carry a CiTO property is a domain stretch; placing the discourse relation on the body resource (which is itself an addressable RDF resource) is a cleaner alternative.

Argument Model Ontology

http://purl.org/spar/amo

Summary: AMO is a SPAR ontology that encodes Toulmin's model of argument. It defines six interrelated components (claim, evidence, warrant, backing, qualifier, rebuttal) organized as roles within an amo:Argument container. The first three (claim, evidence, warrant) are mandatory components of an argument; the others are optional. AMO is aligned with CiTO, which this model already uses.

Core terms:

• amo:Argument (the container)
• amo:ArgumentationEntity (umbrella for the component types)
• amo:Claim, amo:Evidence, amo:Warrant, amo:Backing, amo:Qualifier, amo:Rebuttal (the components)
• Object properties: amo:hasClaim, amo:hasEvidence, amo:hasWarrant (and the others), plus inter-component relations like amo:proves (evidence → claim), amo:leadsTo (warrant → claim), amo:backs (backing → warrant), amo:forces (qualifier → claim), amo:isValidUnless (claim → rebuttal)

Limitations / gaps:

The component classes are definitionally tied to an argument container. Every individual typed amo:Claim is by the ontology's semantics the claim of some amo:Argument. Typing a content node amo:Claim would commit us to a Toulmin-structured argument the content node may not be part of.

The natural-language scope is also narrower than what we need. amo:Claim is "a fact that must be established," operationally the thing a Toulmin argument is trying to prove. Our use case is broader: any propositional statement contributes to the discourse graph, whether or not it is the conclusion of an argument.

• AMO's amo:Qualifier is a simple degree-of-certainty modifier ("certainly," "possibly," "presumably"), without support for more complex cases such as linking caveats.
• AMO's amo:Rebuttal is a restriction on a claim's validity ("unless X happens"). It is narrower than the everyday "rebuttal" usage and not directly equivalent to a cito:disputes reply.
• AMO remains a useful candidate for pilot communities that want to structure contributions according to Toulmin's framework. See the spec's Extensions for richer structure for the extension.

SEPIO - Scientific Evidence and Provenance Information Ontology

https://github.com/monarch-initiative/SEPIO-ontology/wiki/The-SEPIO-Core-Ontology

Summary: SEPIO is a domain ontology developed by the Monarch Initiative for representing scientific evidence and its provenance. It targets graded support across multiple independent evidence items, conflicting lines of evidence on a shared proposition, and chained provenance trails. SEPIO was developed in alignment with the Global Alliance for Genomics and Health (GA4GH) information model and is used in clinical and biomedical evidence-curation infrastructure.

Core terms:

• Assertion
• Proposition
• Evidence Line
• Evidence Item
• Contribution

Limitations/gaps:

SEPIO's structural topology is fixed: assertions, propositions, evidence lines, and evidence items each play specific roles. Adopting SEPIO as the core would force every contribution into this topology, which is heavier than most Nanoarguments' use cases need.

The published OWL ontology has also drifted from the GA4GH-aligned information model that SEPIO's documentation references, so adoption would require pinning to a specific version.

PROV-O (W3C Rec)

https://www.w3.org/TR/prov-o/

Summary: PROV-O is the W3C-recommended OWL ontology for representing provenance. It captures who generated what, when, from what source, through what activity. The core triad is prov:Entity, prov:Activity, and prov:Agent. Attribution, derivation, generation, association, and informing relations connect them. PROV-O is widely deployed across scientific data publishing, including the nanopublication ecosystem, where it is the canonical vocabulary for the provenance graph.

Core terms:

• prov:Entity, prov:Activity, prov:Agent
• prov:wasGeneratedBy, prov:used, prov:wasAttributedTo, prov:wasDerivedFrom, prov:wasAssociatedWith, prov:wasInformedBy

Limitations/gaps:

PROV-O is generic by design. Its relations cover provenance broadly without committing to domain-specific notions of derivation, evaluation, or implementation. The Nanoarguments model uses PROV-O for what it covers (attribution, derivation, generation) and mints na:tests and na:implements for the domain-specific relations PROV-O doesn't aim to provide (Relations on statements).

The nanopublication network uses PROV-O conventions specific to its own structure; this specification follows those conventions (Provenance and authorship) per the nanopublication guidelines.

CiTO - Citation Typing Ontology

https://sparontologies.github.io/cito/current/cito.html

Summary: CiTO is a SPAR ontology that types citations and references between scholarly works (and, more generally, between any two resources where one references the other). It defines a flat set of citation-typing properties grouped into positive, negative, neutral, and factual categories. CiTO is widely deployed across the scholarly publishing ecosystem and is one of the most commonly reused vocabularies for typed scholarly relationships.

Core terms:

• cito:cites / cito:isCitedBy
• positive: cito:supports, cito:confirms, cito:extends
• negative: cito:disputes, cito:refutes, cito:critiques, cito:disagreesWith
• neutral: cito:reviews, cito:discusses, cito:citesAsAuthority
• factual: cito:usesMethodIn, cito:usesDataFrom, cito:citesAsDataSource
• cito:Citation + cito:hasCitingEntity / cito:hasCitationCharacterization / cito:hasCitedEntity

Limitations/gaps:

CiTO was primarily designed to type citations between scholarly works (document/work-level citation relationships). It can represent statement-level citation and support relations, but statement-centric knowledge graphs are not its principal design focus. Systems such as dokieli use it at statement or fragment granularity by treating statements as addressable RDF resources and applying CiTO relations between them.

ECO - Evidence & Conclusion Ontology

http://evidenceontology.org/

Summary: ECO is a community-curated controlled vocabulary of evidence categories used to describe the type of evidence supporting a scientific assertion. It provides a structured hierarchy of evidence types crossed with assertion methods, producing leaf terms like "experimental evidence used in manual assertion" or "computational evidence used in automatic assertion." ECO is widely deployed in biomedical resources including the Gene Ontology and UniProt.

Core terms:

• evidence (root class)
• assertion method (root class - manual assertion / automatic assertion)
• evidence x assertion method cross-product leaf terms

Limitations/gaps:

ECO types evidence; it does not type contributions more broadly. It complements rather than replaces other vocabularies in this survey. A contribution typed sio:evidence may optionally carry an ECO class as a finer-grained subtype (Optional evidence typing with ECO). ECO is not required by the model.

SIO - Semanticscience Integrated Ontology

https://github.com/MaastrichtU-IDS/semanticscience

Summary: SIO is a broad upper-level ontology for the description of scientific objects, processes, and information. It provides class hierarchies for propositions, evidence, hypotheses, arguments, processes, and roles, along with general-purpose relations. SIO sits above more specific domain ontologies and provides shared abstractions across biology, chemistry, and other sciences.

Core terms:

• sio:argument
• sio:evidence
• sio:justification

Limitations/gaps:

SIO provides the class hierarchy for hypothesis, evidence, argument, and proposition. This model uses SIO classes for typing and supplies the relations among them through CiTO and project-minted predicates (Relations on statements, SIO subtypes).

ActivityPub

https://www.w3.org/TR/activitypub/

Summary: ActivityPub is a W3C-recommended decentralized social networking protocol built on Activity Streams 2.0. It defines a client-to-server API (clients post to a user's outbox) and a server-to-server federation protocol (servers deliver activities to recipients' inboxes). Each actor has an inbox and an outbox.

Core terms:

• Actor (inbox / outbox)
• delivery (outbox -> inbox federation)

Limitations/gaps:

ActivityPub is a federation transport, not a content vocabulary. This model's content lives in nanopublications, and the connector layer projects nanopublications into ActivityPub activities when federation is needed. ActivityPub and nanopublications have different commitments (federated messaging vs signed immutable graphs), and the connector cannot be fully faithful to both (Alignment with ActivityPub).

Activity Streams 2.0 (alignment target)

https://www.w3.org/TR/activitystreams-vocabulary/

Summary: Activity Streams 2.0 is a W3C-recommended vocabulary for describing social activities and objects on the web. It defines a small set of activity types (Create, Update, Like, Follow, and others) and object types (Note, Article, etc.), along with relations like as:inReplyTo (threading), as:object, and as:actor. It is the data model underlying ActivityPub federation.

Core terms:

• as:Activity (and activity subtypes: Create, Update, etc.)
• as:Note
• as:Object, as:actor, as:object
• as:inReplyTo

Limitations / gaps:

Activity Streams covers conversational structure and social activities. This model uses as:inReplyTo for threading and uses CiTO predicates for rhetorical typing, treating the two as independent axes (Discourse-graph relations).

Linked Data Notifications

https://www.w3.org/TR/ldn/

Summary: Linked Data Notifications (LDN) is a W3C-recommended protocol for delivering structured notifications to a target resource's inbox. The target declares an ldp:inbox IRI; senders POST notification payloads to it. LDN is transport-only and the payload contents are arbitrary linked data (typically Activity Streams activities or Web Annotations).

Core terms:

• ldp:inbox

Limitations/gaps:

LDN delivers notifications but does not define their semantics. The model treats LDN as one available transport channel for annotation delivery; the model itself doesn't depend on LDN being used.

Discourse-graph relations (Statement to Statement)

These relations have a statement as subject and a statement as object. They form the graph among contributions and are what discourse-graph traversal queries (Level 1: graph-traversal queries) operate on.

Context relations (Statement to non-Statement)

These relations have a statement as subject and a non-statement resource as object, typically an external paper, dataset, method, protocol, or other resource that exists outside the discourse graph.

Relations from other layers

The annotation layer (The annotation layer) introduces annotation properties (oa:hasBody, oa:hasTarget, oa:hasSelector) that sit on oa:Annotation resources rather than on content nodes. These describe anchoring (where a body is positioned in relation to a target) and are distinct from discourse relations in subject type and purpose.

The body of an annotation is itself a content node and may carry the full discourse vocabulary. A body can simultaneously be the body of an oa:Annotation (positioning it at a target) and the subject of cito:supports, cito:disputes, cito:qualifies, scientific relations, or citation triples. The annotation framing does not flatten or restrict the body's discourse content; the body's RDF resource is independently a full participant in the discourse layer.

The discourse layer

The discourse layer lives on content nodes as direct triples. Discourse relations (cito:supports, cito:disputes, cito:qualifies, as:inReplyTo, and other CiTO relations from the subset in section 4) sit on the content node and connect it to its target. This layer carries statements about how a contribution relates to others in the discussion: who supports or disputes what, who replies to whom, who qualifies whose claims.

The discourse layer is always present on a content node that has any discourse standing regardless of the authoring context. It is the canonical representation of the discussion structure and what aggregation queries (Query patterns) operate on.

The annotation layer

The annotation layer uses oa:Annotation to anchor a body (typically a content node) with a target. The annotation is a separate, addressable resource with its own properties: a body, a target, optionally a selector narrowing the target to a part of it, and optionally a motivation. The target may be a document, a content node, or a part of either, expressed where appropriate as an oa:SpecificResource carrying an oa:hasSelector. This layer carries statements about positioning and association: which body is attached where, with what motivation, by whom.

The canonical case is an annotation tool that operates on documents: every tool-authored contribution naturally carries an annotation layer with a document or part of a document as the target (dokieli is one such tool). Annotation targets are not restricted to documents. Any addressable resource can be a target, including other content nodes, allowing one annotation to position a contribution within a target produced elsewhere in the network.

Orthogonality

The two layers are independent:

• A contribution may participate in the discourse layer only. Example: a claim disputing another claim, published directly to the network, with no annotation associating it with a target.
• A contribution may be the body of an annotation only. Example: a comment that associates itself with a passage of a document, without taking any rhetorical position toward another content node.
• A contribution may participate in both. Example: a rebuttal that disputes a counter-argument (discourse layer) and is also associated with a specific passage of that counter-argument via an annotation (annotation layer).

When both layers are present they describe the contribution from different angles and do not duplicate information. The annotation's body and the content node share an IRI as they are the same RDF resource. The triples about that resource appear once with discourse relations stated directly on it, and the annotation references it as oa:hasBody. Anchoring stays on the annotation while the rhetorical relations stay on the content node.

Layers and nanopublications

The two layers may coexist in a single nanopublication when they represent one authoring act. A dokieli user who selects a passage and writes a rebuttal in one operation produces a single nanopub whose assertion graph contains both the rebuttal content node with its discourse triples and the annotation anchoring the body to the passage.

The two layers occupy different nanopublications when they represent separate acts, for example, a content node published in one nanopub by one author, later anchored to a passage by an annotation in a separate nanopub by a different author or at a different time.

Interoperability between the two shapes

The two representations are alternate views of the same underlying fact, and are bidirectionally derivable at query time.

Assertion authorship (provenance graph)

The provenance graph records who is responsible for the content of the assertion and where the assertion came from:

sub:assertion prov:wasAttributedTo <agent-IRI> ;
  prov:wasDerivedFrom <source-IRI> .

<agent-IRI> is an ORCID for a human author or a stable IRI for a non-human agent. <source-IRI> (where applicable) identifies the source the assertion was derived from (a paper, dataset, experiment, upstream nanopub). Optionally, prov:generatedAtTime records when the assertion was made.

The provenance graph MUST contain at least one triple referencing the assertion. prov:wasAttributedTo is the recommended way to do this when an agent is responsible for the assertion's content.

When the assertion graph contains an oa:Annotation resource, the annotation's authorship is by default the same as the assertion's: a single prov:wasAttributedTo on the assertion attributes the assertion's content, including any annotation resources within it, to one agent. If the annotation's authorship differs from the assertion's, the annotation resource carries its own attribution, using oa:creator (or dct:creator) directly on the annotation. The common case where this matters is third-party annotation: a curator publishes a nanopub whose assertion contains a pre-existing claim by Author A and an annotation by the curator associating that claim with a target.

Nanopub authorship (pubinfo graph)

The pubinfo graph records who is responsible for the act of publishing the nanopub. The canonical convention is as follows:

this: dct:creator <agent-IRI> ;
  dct:created "..."^^xsd:dateTime .

The pubinfo SHOULD contain attribution and a timestamp. The pubinfo graph also carries the cryptographic signature (sub:sig with npx:signedBy, npx:hasPublicKey, npx:hasSignature), the license (dct:license), and any indexing properties such as npx:hasNanopubType.

Assertion creator and nanopub creator

The assertion creator (provenance graph) and the nanopub creator (pubinfo graph) may identify different agents. There are two common patterns:

• Same agent in both graphs: a human publishes their own assertion through an application, or an autonomous bot does both. The same IRI appears in each graph but under different predicates (prov:wasAttributedTo in provenance, dct:creator in pubinfo). This is not a redundancy, as each graph attributes a different act to the same agent.

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0003-1062-5576 .
}

sub:pubinfo {
  this: dct:creator orcid:0000-0003-1062-5576 ;
    dct:created "2026-06-08T08:08:08Z"^^xsd:dateTime .
}

• Different agents in each graph: one agent authored the assertion content; a different agent (a curator, a colleague, an organization, an automated publishing tool) is responsible for the publication act. This pattern lets a publisher republish someone else's assertion without claiming the content as their own, or attribute content to its source while taking responsibility for putting it on the network.

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0003-3934-0072 .
}

sub:pubinfo {
  this: dct:creator orcid:0000-0003-0183-6910 ;
    dct:created "..."^^xsd:dateTime .
}

Tool-mediated publishing

When a tool publishes a nanopub on behalf of a human author, the assertion creator and the nanopub creator may identify different agents (Assertion creator and nanopub creator). Some patterns include:

• Naming the tool as dct:creator of the nanopub directly (the tool is the publisher)
• Keeping the human as dct:creator of the nanopub and naming the tool via a separate property (e.g. npx:wasCreatedAt)
• Recording the originating session, template, or input as prov:wasDerivedFrom

This specification does not commit to a single pattern, and follows whatever the publishing tool emits. Queries that need to identify the tool that produced a given nanopub can look for any of these properties. A canonical emission convention may be added in a later version once the implementation milestone has chosen one.

Agents

Agents are identified by IRIs. A human is identified by an ORCID; a bot, tool, or automated agent is identified by a stable IRI managed by the project or organization that operates it (e.g. https://w3id.org/kpxl/gen/terms/RoCrateBot ). The model does not require type declaration on agents. An ORCID and a bot IRI are both just creators, treated symmetrically.

Pubinfo may optionally carry a foaf:name on the agent's IRI for display purposes. This is convenience metadata for UIs, not a requirement of the model.

Source application or platform

When a contribution originates from a federated platform or decentralized clientside tool (a discussion forum, an annotation tool, an authoring environment), the originating platform may be recorded in pubinfo so UIs can display source attribution ("via discussit.org") and federation logic can route accordingly. Two properties are seen for this purpose:

• npx:wasCreatedAt: used by Nanodash to identify the publishing tool.
• oa:renderedVia: Web Annotation property used by dokieli to identify the tool an annotation was rendered through.

Their exact semantic distinction is not documented in this specification. This document treats either as a source of contribution marker for query and display purposes. The choice of which property to emit is left to the publishing tool based on their specific context.

Nanopub type indexing

The pubinfo graph carries npx:hasNanopubType indicating the primary rhetorical or content type of the nanopub (e.g. cito:disputes, schema:Statement, cito:qualifies). Used for indexing and query.

SIO subtypes

The following SIO classes may be used alongside schema:Statement to refine the contribution's intrinsic type.

• sio:hypothesis (SIO_000284): "a proposed explanation for a phenomenon."
• sio:evidence (SIO_001394): "the body of facts or information indicating whether a belief or proposition is true or valid."
• sio:argument (SIO_000259): "a set of one or more declarative sentences (or propositions) known as the premises along with another declarative sentence (or proposition) known as the conclusion."

All three are subclasses of sio:proposition (SIO_000256), defined as "a sentence expressing something true or false." SIO does not define structural relationships between these concepts; relating them is this model's job, via the discourse-graph relations (Discourse-graph relations).

These types are intrinsic. A statement is sio:evidence because of what it is (a body of facts bearing on a proposition), not because of which claim it currently grounds. What is contextual is the role the statement plays in a particular argument, and that role is expressed by the discourse-graph relations connecting it to other statements.

Additional SIO classes (e.g. for studies, findings, conclusions, beliefs) may be added as pilot use cases surface specific needs.

Optional evidence typing with ECO

A statement typed sio:evidence (or a domain-specific subclass) may optionally carry a finer-grained evidence type drawn from the Evidence and Conclusion Ontology (ECO). ECO is a community-curated controlled vocabulary of evidence categories ("experimental evidence," "computational evidence," "author statement used in manual assertion," etc.), widely used by biomedical resources including the Gene Ontology and UniProt.

ECO typing is structural-model-neutral: it adds a dct:type (or equivalent) pointing at an ECO class on an existing statement and does not change this model's topology.

Example: a finding typed as evidence, with the SIO class for the intrinsic type and an ECO class for the leaf category:

sub:finding a sio:evidence ;
  rdf:value     "Pigeons pecked at mirror-visible stickers but not at hidden controls." ;
  dct:type      eco:0000006 ;  # experimental evidence
  cito:supports sub:hypothesis .

A receiver that understands neither SIO nor ECO still sees a statement with a value and a cito:supports relation; the typing degrades gracefully.

Extensions for richer structure

Two existing ontologies provide structure this model's core does not commit to, but which some pilot communities may want. Each is available as an extension.

• SEPIO: richer evidence-line semantics. Where a contribution needs to group multiple findings into a single coherent line of support or refutation, the recommended representation is an intermediate statement that the findings target via cito:supports and that itself targets the claim. This handles most evidence-grouping cases without additional machinery. For use cases that require richer semantics — graded support across multiple independent lines, chained provenance trails, contradicting lines on a shared proposition, or alignment with existing SEPIO-based infrastructure — SEPIO (see Part I) may be adopted as an extension. Adoption would require pinning to a specific version (the published OWL ontology has drifted from the GA4GH-aligned information model).
• AMO: Toulmin-structured arguments. Where a contribution should be structured according to Toulmin's model of argument, distinguishing claim, evidence, warrant, backing, qualifier, and rebuttal as named components of an explicit argument, the Argument Model Ontology (AMO) may be adopted. AMO is a SPAR ontology aligned with CiTO, with first-class classes for each Toulmin component and object properties for the relations between them. This model's core does not adopt AMO's component classes because they are definitionally tied to an amo:Argument container (amo:Claim is equivalently defined as the claim of some argument) and the core treats statements as standing on their own without commitment to a Toulmin framework.

Level 1: graph-traversal queries

Level 1 queries follow typed edges between content nodes, with no aggregation, and equivalent claims are not merged. The pattern supports retrieving a contribution's incoming and outgoing relations by type, walking reply or rhetorical chains to arbitrary depth, and resolving each contribution's author and pubinfo metadata. Each contribution is returned as a distinct result row; equivalent contributions by different agents are not merged.

Find all contributions targeting a given claim, with their relation type:

SELECT ?relation ?contribution ?author WHERE {
  ?contribution ?relation <CLAIM_IRI> ;
                ^prov:wasAttributedTo ?author .
  FILTER (?relation IN (cito:supports, cito:disputes,
                        cito:qualifies, cito:agreesWith,
                        cito:extends, as:inReplyTo))
}

Walk a reply tree from a root:

SELECT ?contribution ?parent ?relation ?author WHERE {
  ?contribution as:inReplyTo+ <ROOT_CLAIM_IRI> ;
                as:inReplyTo  ?parent ;
                ^prov:wasAttributedTo ?author .
  OPTIONAL {
    ?contribution ?relation ?parent .
    FILTER (?relation IN (cito:supports, cito:disputes,
                          cito:qualifies, cito:agreesWith,
                          cito:extends))
  }
}

Find scientific reasoning chains:

SELECT ?contribution ?type ?predecessor ?relation WHERE {
  ?contribution (prov:wasDerivedFrom|na:tests|na:implements)+ <HYPOTHESIS_IRI> ;
                a ?type ;
                ?relation ?predecessor .
  FILTER (?relation IN (prov:wasDerivedFrom, na:tests, na:implements))
  FILTER (?type IN (sio:hypothesis, sio:evidence, sio:argument, schema:Statement))
}

Implementations are expected to provide query templates of this shape in the templates and query patterns milestone of the Nanoarguments project.

Level 2: aggregation queries (out of scope for this version)

Aggregation queries derive summary information by grouping, counting, ranking, or otherwise combining results across multiple contributions. Categories include:

• Equivalent-claim sets.
• Evidence aggregation (grouped lines of evidence, graded support, conflict resolution).
• Agent-level aggregation (per-author counts, agreement clusters, derived signals).

These queries depend on vocabulary and convention additions that this spec does not commit to. They are out of scope and deferred to a follow-up specification effort.

Direct discourse representation

Each contribution is a nanopublication. Discourse relations are direct triples on the content nodes.

The thread consists of:

• A root claim about gut microbiota and vaccine response.
• A supporting contribution.
• A caveat noting a methodological concern.
• A counter-argument disputing the root claim.
• A rebuttal of the counter-argument, citing evidence.
• An open question requesting clarification.

@prefix this:   <https://w3id.org/np/RAxample-alvarez-microbiota-claim> .
@prefix sub:    <https://w3id.org/np/RAxample-alvarez-microbiota-claim/> .
@prefix np:     <http://www.nanopub.org/nschema#> .
@prefix npx:    <http://purl.org/nanopub/x/> .
@prefix prov:   <http://www.w3.org/ns/prov#> .
@prefix dct:    <http://purl.org/dc/terms/> .
@prefix xsd:    <http://www.w3.org/2001/XMLSchema#> .
@prefix rdf:    <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix orcid:  <https://orcid.org/> .
@prefix schema: <https://schema.org/> .
@prefix cito:   <http://purl.org/spar/cito/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:claim a schema:Statement ;
    rdf:value             "Certain gut microbiota profiles have stronger mRNA vaccine responses." ;
    cito:citesAsEvidence  <https://doi.org/10.1234/kim-2025-microbiota-vaccine> .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0001-alvarez-b ;
    prov:generatedAtTime "2026-05-24T08:00:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0001-alvarez-b ;
    dct:created            "2026-05-24T08:00:00Z"^^xsd:dateTime ;
    npx:introduces       sub:claim ;
    npx:hasNanopubType     schema:Statement .
}

@prefix this: <https://w3id.org/np/RAxample-wang-support> .
@prefix sub:  <https://w3id.org/np/RAxample-wang-support/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:reason a schema:Statement ;
    rdf:value      "Their SCFA data is compelling." ;
    cito:supports  <https://w3id.org/np/RAxample-alvarez-microbiota-claim/claim> .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0002-wang-p ;
    prov:generatedAtTime "2026-05-24T09:00:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0002-wang-p ;
    dct:created            "2026-05-24T09:00:00Z"^^xsd:dateTime ;
    npx:introduces       sub:reason ;
    npx:hasNanopubType     cito:supports .
}

@prefix this: <https://w3id.org/np/RAxample-medaibot-caveat> .
@prefix sub:  <https://w3id.org/np/RAxample-medaibot-caveat/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:caveat a schema:Statement ;
    rdf:value     "Be cautious about interpreting causality here without intervention studies." ;
    cito:qualifies  <https://w3id.org/np/RAxample-alvarez-microbiota-claim/claim> .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo <https://example.org/agents/medai-bot> ;
    prov:generatedAtTime "2026-05-24T10:10:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        <https://example.org/agents/medai-bot> ;
    dct:created            "2026-05-24T10:10:00Z"^^xsd:dateTime ;
    npx:introduces       sub:caveat ;
    npx:hasNanopubType     cito:qualifies .
}

@prefix this: <https://w3id.org/np/RAxample-moller-counterargument> .
@prefix sub:  <https://w3id.org/np/RAxample-moller-counterargument/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:counterclaim a schema:Statement ;
    rdf:value      "The correlation could be confounded by diet." ;
    cito:disputes  <https://w3id.org/np/RAxample-alvarez-microbiota-claim/claim> .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0003-moller-m ;
    prov:generatedAtTime "2026-05-24T10:28:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0003-moller-m ;
    dct:created            "2026-05-24T10:28:00Z"^^xsd:dateTime ;
    npx:introduces       sub:counterclaim ;
    npx:hasNanopubType     cito:disputes .
}

@prefix this: <https://w3id.org/np/RAxample-criticaibot-rebuttal> .
@prefix sub:  <https://w3id.org/np/RAxample-criticaibot-rebuttal/> .
@prefix as:   <https://www.w3.org/ns/activitystreams#> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:rebuttal a schema:Statement ;
    rdf:value             "Note that [Kim 2025] controlled for fiber intake using dietary logs." ;
    cito:citesAsEvidence  <https://doi.org/10.1234/kim-2025-microbiota-vaccine> ;
    as:inReplyTo          <https://w3id.org/np/RAxample-moller-counterargument/counterclaim> ;
    cito:disputes         <https://w3id.org/np/RAxample-moller-counterargument/counterclaim> .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo <https://example.org/agents/criticai-bot> ;
    prov:generatedAtTime "2026-05-24T10:48:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        <https://example.org/agents/criticai-bot> ;
    dct:created            "2026-05-24T10:48:00Z"^^xsd:dateTime ;
    npx:introduces       sub:rebuttal ;
    npx:hasNanopubType     cito:disputes .
}

@prefix this: <https://w3id.org/np/RAxample-perez-question> .
@prefix sub:  <https://w3id.org/np/RAxample-perez-question/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:question a schema:Question ;
    rdf:value     "Has anyone tested this in non-trial populations?" ;
    as:inReplyTo  <https://w3id.org/np/RAxample-alvarez-microbiota-claim/claim> .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0002-perez-l ;
    prov:generatedAtTime "2026-05-24T11:15:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0002-perez-l ;
    dct:created            "2026-05-24T11:15:00Z"^^xsd:dateTime ;
    npx:introduces       sub:question ;
    npx:hasNanopubType     schema:Question .
}

Annotation-based representation

The same six contributions, expressed as a thread where each contribution after the root is published as an oa:Annotation targeting a previous contribution, as an annotation tool would emit. Bodies are content nodes carrying the same discourse relations as in Direct discourse representation; the annotation wrapper adds anchoring. The root claim (Root claim) is the same nanopublication shown in Root claim, with no annotation.

A note on targets: an annotation can target any addressable resource (a document, a passage of a document, an image region, another annotation, a content node, or any other URI). The choice depends on the context and use case. In the examples below, the contributions are annotations of a claim published as a nanopublication, so each annotation's oa:hasTarget is the canonical nanopublication IRI (https://w3id.org/np/...). This anchoring is distinct from the discourse relation: the oa:hasTarget records where the annotation is positioned (the nanopublication as a citeable artifact), while the discourse relation on the content node (cito:supports, cito:qualifies, as:inReplyTo, etc.) points at the content node being responded to (the /claim resource), matching the direct representation in Direct discourse representation. The two may legitimately point at different granularities (see Counter-argument, where the discourse relation targets the whole claim while the anchoring narrows to a phrase).

@prefix this: <https://w3id.org/np/RAxample-wang-support-annotation> .
@prefix sub:  <https://w3id.org/np/RAxample-wang-support-annotation/> .
@prefix oa:   <http://www.w3.org/ns/oa#> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:reason a schema:Statement ;
    rdf:value      "Their SCFA data is compelling." ;
    cito:supports  <https://w3id.org/np/RAxample-alvarez-microbiota-claim/claim> .

  sub:annot a oa:Annotation ;
    oa:hasBody     sub:reason ;
    oa:hasTarget   <https://w3id.org/np/RAxample-alvarez-microbiota-claim> ;
    oa:motivatedBy oa:assessing .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0002-wang-p ;
    prov:generatedAtTime "2026-05-24T09:00:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0002-wang-p ;
    dct:created            "2026-05-24T09:00:00Z"^^xsd:dateTime ;
    npx:introduces       sub:reason ;
    npx:hasNanopubType     cito:supports .
}

@prefix this: <https://w3id.org/np/RAxample-medaibot-caveat-annotation> .
@prefix sub:  <https://w3id.org/np/RAxample-medaibot-caveat-annotation/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:caveat a schema:Statement ;
    rdf:value     "Be cautious about interpreting causality here without intervention studies." ;
    cito:qualifies  <https://w3id.org/np/RAxample-alvarez-microbiota-claim/claim> .

  sub:annot a oa:Annotation ;
    oa:hasBody     sub:caveat ;
    oa:hasTarget   <https://w3id.org/np/RAxample-alvarez-microbiota-claim> ;
    oa:motivatedBy oa:assessing .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo <https://example.org/agents/medai-bot> ;
    prov:generatedAtTime "2026-05-24T10:10:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        <https://example.org/agents/medai-bot> ;
    dct:created            "2026-05-24T10:10:00Z"^^xsd:dateTime ;
    npx:introduces       sub:caveat ;
    npx:hasNanopubType     cito:qualifies .
}

@prefix this: <https://w3id.org/np/RAxample-moller-counterargument-annotation> .
@prefix sub:  <https://w3id.org/np/RAxample-moller-counterargument-annotation/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:counterclaim a schema:Statement ;
    rdf:value      "The correlation could be confounded by diet." ;
    cito:disputes  <https://w3id.org/np/RAxample-alvarez-microbiota-claim/claim> .

  sub:disputedPhrase a oa:SpecificResource ;
    oa:hasSource <https://w3id.org/np/RAxample-alvarez-microbiota-claim> ;
    oa:hasSelector [ a oa:TextQuoteSelector ;
                     oa:exact "stronger mRNA vaccine responses" ] .

  sub:annot a oa:Annotation ;
    oa:hasBody     sub:counterclaim ;
    oa:hasTarget   sub:disputedPhrase ;
    oa:motivatedBy oa:assessing .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0003-moller-m ;
    prov:generatedAtTime "2026-05-24T10:28:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0003-moller-m ;
    dct:created            "2026-05-24T10:28:00Z"^^xsd:dateTime ;
    npx:introduces       sub:counterclaim ;
    npx:hasNanopubType     cito:disputes .
}

@prefix this: <https://w3id.org/np/RAxample-criticaibot-rebuttal-annotation> .
@prefix sub:  <https://w3id.org/np/RAxample-criticaibot-rebuttal-annotation/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:rebuttal a schema:Statement ;
    rdf:value             "Note that [Kim 2025] controlled for fiber intake using dietary logs." ;
    cito:citesAsEvidence  <https://doi.org/10.1234/kim-2025-microbiota-vaccine> ;
    as:inReplyTo          <https://w3id.org/np/RAxample-moller-counterargument-annotation/counterclaim> ;
    cito:disputes         <https://w3id.org/np/RAxample-moller-counterargument-annotation/counterclaim> .

  sub:annot a oa:Annotation ;
    oa:hasBody     sub:rebuttal ;
    oa:hasTarget   <https://w3id.org/np/RAxample-moller-counterargument-annotation> ;
    oa:motivatedBy oa:replying .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo <https://example.org/agents/criticai-bot> ;
    prov:generatedAtTime "2026-05-24T10:48:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        <https://example.org/agents/criticai-bot> ;
    dct:created            "2026-05-24T10:48:00Z"^^xsd:dateTime ;
    npx:introduces       sub:rebuttal ;
    npx:hasNanopubType     cito:disputes .
}

@prefix this: <https://w3id.org/np/RAxample-perez-question-annotation> .
@prefix sub:  <https://w3id.org/np/RAxample-perez-question-annotation/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:question a schema:Question ;
    rdf:value     "Has anyone tested this in non-trial populations?" ;
    as:inReplyTo  <https://w3id.org/np/RAxample-alvarez-microbiota-claim/claim> .

  sub:annot a oa:Annotation ;
    oa:hasBody     sub:question ;
    oa:hasTarget   <https://w3id.org/np/RAxample-alvarez-microbiota-claim> ;
    oa:motivatedBy oa:questioning .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0002-perez-l ;
    prov:generatedAtTime "2026-05-24T11:15:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0002-perez-l ;
    dct:created            "2026-05-24T11:15:00Z"^^xsd:dateTime ;
    npx:introduces       sub:question ;
    npx:hasNanopubType     schema:Question .
}

Equivalence of the two representations

The content nodes in Direct discourse representation and Annotation-based representation are the same: same IRIs, same values, same discourse-graph relations. A query for "all contributions targeting Alvarez's claim" returns the same result set in either representation, because the cito:supports, cito:disputes, cito:qualifies, and as:inReplyTo triples sit on the content nodes regardless of whether an annotation wraps them.

What differs between the two representations is the anchoring: Annotation-based representation contributions additionally carry oa:Annotation resources with oa:hasTarget pointing at the contribution they respond to. Direct discourse representation contributions have no such anchoring; the discourse-graph relation is the only structural connection.

For a network mixing both representations (some contributions published as plain nanopubs, others as annotations), the discourse-graph traversal queries of Level 1: graph-traversal queries operate uniformly. Implementations interested in the additional anchoring information can additionally query oa:hasTarget to recover where annotations are positioned.

Discourse rooted in an annotation of a paper

The previous examples started with a claim published as a nanopublication. A different common pattern starts with an annotation of an external document: a contributor reads a paper, selects a passage, and writes a claim about it. Subsequent contributions respond by annotating the original annotation, producing a chain where each annotation's target is the previous one.

This section uses a different topic from Direct discourse representation and Annotation-based representation to keep the patterns distinct. The thread:

• A root annotation of a passage in a paper, asserting a counter-claim.
• A support annotation of the root, citing evidence.
• A counter-argument annotation of the root.
• A rebuttal annotation of the counter-argument, citing further evidence.

@prefix this:   <https://w3id.org/np/RAxample-pigeon-mirror-root> .
@prefix sub:    <https://w3id.org/np/RAxample-pigeon-mirror-root/> .
@prefix np:     <http://www.nanopub.org/nschema#> .
@prefix npx:    <http://purl.org/nanopub/x/> .
@prefix prov:   <http://www.w3.org/ns/prov#> .
@prefix dct:    <http://purl.org/dc/terms/> .
@prefix xsd:    <http://www.w3.org/2001/XMLSchema#> .
@prefix rdf:    <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix orcid:  <https://orcid.org/> .
@prefix schema: <https://schema.org/> .
@prefix oa:     <http://www.w3.org/ns/oa#> .
@prefix cito:   <http://purl.org/spar/cito/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:claim a schema:Statement ;
    rdf:value      "Pigeons can pass a modified mirror-mark test." ;
    cito:disputes  <https://example.org/chimpanzee-self-recognition> .

  sub:annot a oa:Annotation ;
    oa:hasBody     sub:claim ;
    oa:hasTarget   [ a oa:SpecificResource ;
                     oa:hasSource <https://example.org/chimpanzee-self-recognition> ;
                     oa:hasSelector [ a oa:TextQuoteSelector ;
                                      oa:exact  "no nonhuman species other than the great apes" ;
                                      oa:prefix "We conclude that " ;
                                      oa:suffix " demonstrates self-recognition." ] ] ;
    oa:motivatedBy oa:assessing .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0002-1234-5678 ;
    prov:generatedAtTime "2026-05-24T11:05:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0002-1234-5678 ;
    dct:created            "2026-05-24T11:05:03Z"^^xsd:dateTime ;
    npx:introduces       sub:claim ;
    npx:hasNanopubType     cito:disputes .
}

@prefix this: <https://w3id.org/np/RAxample-pigeon-mirror-support> .
@prefix sub:  <https://w3id.org/np/RAxample-pigeon-mirror-support/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:evidence a schema:Statement ;
    rdf:value             "Trained pigeons pecked at body stickers only visible via a mirror, not at hidden controls." ;
    cito:citesAsEvidence  <https://doi.org/10.1111/pigeon-mirror-mark-test> ;
    cito:supports         <https://w3id.org/np/RAxample-pigeon-mirror-root/claim> .

  sub:annot a oa:Annotation ;
    oa:hasBody     sub:evidence ;
    oa:hasTarget   <https://w3id.org/np/RAxample-pigeon-mirror-root/annot> ;
    oa:motivatedBy oa:assessing .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0003-2345-6789 ;
    prov:generatedAtTime "2026-05-24T12:30:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0003-2345-6789 ;
    dct:created            "2026-05-24T12:30:00Z"^^xsd:dateTime ;
    npx:introduces       sub:evidence ;
    npx:hasNanopubType     cito:supports .
}

@prefix this: <https://w3id.org/np/RAxample-pigeon-mirror-counter> .
@prefix sub:  <https://w3id.org/np/RAxample-pigeon-mirror-counter/> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:counterclaim a schema:Statement ;
    rdf:value      "The training conditions may not generalize beyond the lab setting." ;
    cito:disputes  <https://w3id.org/np/RAxample-pigeon-mirror-root/claim> .

  sub:annot a oa:Annotation ;
    oa:hasBody     sub:counterclaim ;
    oa:hasTarget   <https://w3id.org/np/RAxample-pigeon-mirror-root/annot> ;
    oa:motivatedBy oa:assessing .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0004-3456-7890 ;
    prov:generatedAtTime "2026-05-24T14:15:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0004-3456-7890 ;
    dct:created            "2026-05-24T14:15:00Z"^^xsd:dateTime ;
    npx:introduces       sub:counterclaim ;
    npx:hasNanopubType     cito:disputes .
}

@prefix this: <https://w3id.org/np/RAxample-pigeon-mirror-rebuttal> .
@prefix sub:  <https://w3id.org/np/RAxample-pigeon-mirror-rebuttal/> .
@prefix as:   <https://www.w3.org/ns/activitystreams#> .

sub:Head {
  this: a np:Nanopublication ;
    np:hasAssertion       sub:assertion ;
    np:hasProvenance      sub:provenance ;
    np:hasPublicationInfo sub:pubinfo .
}

sub:assertion {
  sub:rebuttal a schema:Statement ;
    rdf:value             "Wild populations of corvids show similar mirror-guided behaviour, suggesting the capacity is not lab-specific." ;
    cito:citesAsEvidence  <https://doi.org/10.1016/magpie-mirror-self-recognition-review> ;
    as:inReplyTo          <https://w3id.org/np/RAxample-pigeon-mirror-counter/counterclaim> ;
    cito:disputes         <https://w3id.org/np/RAxample-pigeon-mirror-counter/counterclaim> .

  sub:annot a oa:Annotation ;
    oa:hasBody     sub:rebuttal ;
    oa:hasTarget   <https://w3id.org/np/RAxample-pigeon-mirror-counter/annot> ;
    oa:motivatedBy oa:replying .
}

sub:provenance {
  sub:assertion prov:wasAttributedTo orcid:0000-0005-4567-8901 ;
    prov:generatedAtTime "2026-05-24T15:45:00Z"^^xsd:dateTime .
}

sub:pubinfo {
  this: dct:creator        orcid:0000-0005-4567-8901 ;
    dct:created            "2026-05-24T15:45:00Z"^^xsd:dateTime ;
    npx:introduces       sub:rebuttal ;
    npx:hasNanopubType     cito:disputes .
}