Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 81 of the 81 references it cites were found in a registry or an index; none of the 101 sources its claims cite is checked for support yet.

Rosetta Statements: simplifying FAIR knowledge graph construction with a user-centred approach

Open access

Abstract

Knowledge graphs and ontologies are promising technologies for achieving FAIR (findable, accessible, interoperable, and reusable) data. We identify four challenges as high barriers for the effective use of knowledge graphs. Since the construction of knowledge graphs is a modelling task and every model serves a purpose against which it is optimized, we question the central paradigm of modelling a mind-independent reality. Instead, we propose the Rosetta Statement approach, which models English natural language statements and displays them as natural language sentences in its user interface. We suggest a Resource Description Framework (RDF)-native metamodel, from which semantic data schemata can be derived for any type of simple English statement. We provide a light and a full version, with the latter supporting versioning and a change-track. We implemented the full version in the Open Research Knowledge Graph (ORKG), an open, domain-agnostic, community-driven knowledge graph for documenting research findings from scholarly publications. The ORKG allows domain experts, with short training but without formal expertise in semantics, to define RDF schemata for new types of Rosetta Statements. We discuss how the Rosetta Statement approach contributes to addressing the four challenges and how its structural proximity to natural language supports the development of tools for data entry and summarization using Large Language Models. We further discuss how the Rosetta approach supports a three-step procedure for FAIR knowledge graph construction: (1) domain experts using Rosetta Statements and Wikidata terms to create a FAIR knowledge graph with basic search functionality; (2) the addition of semantic search capability by replacing Wikidata terms with ontology terms; and (3) the transformation of selected Rosetta Statement types into reasoning-capable graphs with support from ontology engineers. We argue that this three-step procedure is designed to substantially lower the entry barrier for knowledge graph construction while increasing their cognitive interoperability, consistent with the CLEAR Principle.

Overview

CoChat Summary

TL;DR

The paper proposes Rosetta Statements, an RDF-native metamodel that models simple English natural-language statements to lower barriers to FAIR knowledge graph construction, with light and full versioning variants implemented in the Open Research Knowledge Graph (ORKG). It argues this user-centred approach improves cognitive interoperability and outlines a three-step path from Wikidata-based statements to ontology-based semantic search and selective reasoning via schema cross-walks.

Key findings

  • The paper identifies four barriers to knowledge graph use: cognitive interoperability, graph querying, semantic parsing burden, and dynamic knowledge graph construction; it proposes modelling natural language statements rather than a mind-independent reality.
  • It defines a light and a full RDF Rosetta Statement metamodel; the full version supports versioning and editing history and is implemented in ORKG, where domain experts can define new statement schemata after short training—preliminary workshops suggest a half-day guided session.
  • In the ORKG example, a measurement statement with confidence interval would require 15 clicks in a semantic data browser but is displayed as a natural language statement by a single click; the Rosetta schema uses three triples instead of five or six in the compared OBO/OBOE examples.
  • It proposes a three-step procedure: (1) domain experts use Rosetta Statements and Wikidata terms to create a FAIR knowledge graph with basic search; (2) replace Wikidata terms with ontology terms for semantic search; (3) transform selected statement types into reasoning-capable graphs with ontology engineers via schema cross-walks.

Methods

  • Conceptual and design paper: introduces the Rosetta Statement approach, an RDF-native metamodel, light and full versions, display templates, and schema cross-walks; no controlled empirical study is reported.
  • Implementation: the full version is deployed in ORKG, an open, domain-agnostic, community-driven knowledge graph; the ORKG type editor creates Rosetta Statement schemata and SHACL shapes, dynamic labels display statements, and statements can be published as nanopublications.
  • Evidence and analysis: design rationale, worked examples (measurement and travel statements), comparisons to OBI, OBOE, QUDT, and relational schemata, plus qualitative feedback from ORKG users and workshops; no quantitative sample, statistical analysis, or user-study measures are reported.

Limitations

  • Stated: the approach has not undergone a controlled empirical evaluation of usability, learnability, or efficiency compared with alternative knowledge graph construction approaches; claims rest on design rationale, ORKG deployment, and qualitative feedback, and systematic user studies are future work.
  • Stated: the current implementation is English-centric; adapting to non-English languages may require modifications for different syntactic orders or morphologically rich structures, and multilingual templates are future work.
  • Stated: the approach prioritizes cognitive interoperability over reasoning, so initial Rosetta graphs are not fully reasoning-capable; reasoning requires schema cross-walks to OWL-based schemata, which require semantic expertise and can be nontrivial. The paper also states that the proportion of scientific knowledge expressible as Rosetta Statements is not quantified and that schema proliferation or ambiguity is not eliminated.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1093/database/baag030

Cites

81 of 81 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Comparison of a human-readable statement with its machine-actionable and its human-actionable representation. Top: A human-readable statement about the observation that a particular apple weighs 241.68 g, with a 95% confidence interval of 241.31–242.05 g. Middle: A machine-actionable representation of the same statement as an ABox semantic graph, using RDF syntax and following the general schema for measurement data from the Ontology for Biomedical Investigations [83] of the Open Biological and Biomedical Ontology Foundry. Marked is a triple that is not semantically meaningful to a domain expert and thus difficult to comprehend for them. Bottom: A human-actionable representation of the same statement as a mind-map like graph, reducing the complexity of the RDF graph to the information that is actually relevant to a human reader. Figure adapted from [18].
Figure 1 Comparison of a human-readable statement with its machine-actionable and its human-actionable representation. Top: A human-readable statement about the observation that a particular apple weighs 241.68 g, with a 95% confidence interval of 241.31–242.05 g. Middle: A machine-actionable representation of the same statement as an ABox semantic graph, using RDF syntax and following the general schema for measurement data from the Ontology for Biomedical Investigations [83] of the Open Biological and Biomedical Ontology Foundry. Marked is a triple that is not semantically meaningful to a domain expert and thus difficult to comprehend for them. Bottom: A human-actionable representation of the same statement as a mind-map like graph, reducing the complexity of the RDF graph to the information that is actually relevant to a human reader. Figure adapted from [18].CC BY · PubMed Central
Figure 2 Parallels between natural language statements and data schemata. (A) A natural language statement is structured by syntactic and grammatical conventions into syntactic positions of phrases of a syntax tree. (B) The syntax tree corresponding to the natural language statement from (A). (C) The formalized statement of the natural language statement from (A), where each position is associated with a specific semantic role, which can be described by a thematic label. (D) A tabular data schema of the natural language statement from (A). (E) A graphical data schema of the natural language statement from (A). Both data schemata must represent the syntactic positions of the natural language statement as slots, and each slot must specify its associated semantic role as a constraint specification.
Figure 2 Parallels between natural language statements and data schemata. (A) A natural language statement is structured by syntactic and grammatical conventions into syntactic positions of phrases of a syntax tree. (B) The syntax tree corresponding to the natural language statement from (A). (C) The formalized statement of the natural language statement from (A), where each position is associated with a specific semantic role, which can be described by a thematic label. (D) A tabular data schema of the natural language statement from (A). (E) A graphical data schema of the natural language statement from (A). Both data schemata must represent the syntactic positions of the natural language statement as slots, and each slot must specify its associated semantic role as a constraint specification.CC BY · PubMed Central
Figure 3 Cross-walk from one schema to another for a weight measurement statement. The same weight measurement statement is modeled using two different schemata. Top: The weight measurement according to the schema of the Ontology for Biomedical Investigations [83] of the Open Biological and Biomedical Ontology (OBO) Foundry, which is often used in the biomedical domain. Bottom: The same weight measurement according to the schema of the Extensible Observation Ontology (OBOE), which is often used in the ecology community. The arrows indicate the alignment of slots that share the same constraint specification, i.e. the same semantic role. The corresponding semantic roles include the MATERIAL ENTITY, the QUALITY, and the VALUE that has been measured together with its UNIT. The slots and their relationships to one another carry the semantic content that actually conveys the meaning of the weight measurement statement to a human reader. Blue arrows indicate slots with resources as values, and green arrows those with values. Slots with purple borders indicate issues with terminological interoperability: OBO uses an instance of the class ‘pato: weight’, while OBOE, in this example, uses an instance of the class ‘ncit: weight’. However, since ‘pato: weight’ and ‘ncit: weight’ are synonymous terms and can therefore be mapped to establish terminological interoperability between them, the two graphs are semantically interoperable.
Figure 3 Cross-walk from one schema to another for a weight measurement statement. The same weight measurement statement is modeled using two different schemata. Top: The weight measurement according to the schema of the Ontology for Biomedical Investigations [83] of the Open Biological and Biomedical Ontology (OBO) Foundry, which is often used in the biomedical domain. Bottom: The same weight measurement according to the schema of the Extensible Observation Ontology (OBOE), which is often used in the ecology community. The arrows indicate the alignment of slots that share the same constraint specification, i.e. the same semantic role. The corresponding semantic roles include the MATERIAL ENTITY, the QUALITY, and the VALUE that has been measured together with its UNIT. The slots and their relationships to one another carry the semantic content that actually conveys the meaning of the weight measurement statement to a human reader. Blue arrows indicate slots with resources as values, and green arrows those with values. Slots with purple borders indicate issues with terminological interoperability: OBO uses an instance of the class ‘pato: weight’, while OBOE, in this example, uses an instance of the class ‘ncit: weight’. However, since ‘pato: weight’ and ‘ncit: weight’ are synonymous terms and can therefore be mapped to establish terminological interoperability between them, the two graphs are semantically interoperable.CC BY · PubMed Central
Figure 4 From the structure of a natural language statement to the structure of an RDF-based Rosetta Statement schema. (A) A natural language statement (token model) with the predicate has (measurement). (B) The corresponding formalized natural language statement (metamodel), with the syntactic positions and their associated semantic roles highlighted in color. (C) The RDF-based Rosetta Statement schema (metamodel) for the has-measurement statement. Arrows indicate the alignment between positions and slots across the three models.
Figure 4 From the structure of a natural language statement to the structure of an RDF-based Rosetta Statement schema. (A) A natural language statement (token model) with the predicate has (measurement). (B) The corresponding formalized natural language statement (metamodel), with the syntactic positions and their associated semantic roles highlighted in color. (C) The RDF-based Rosetta Statement schema (metamodel) for the has-measurement statement. Arrows indicate the alignment between positions and slots across the three models.CC BY · PubMed Central
Figure 5 From a formalized natural language statement to the corresponding light version of the RDF Rosetta Statement metamodel. (A) A formalized statement with its syntactic positions and associated semantic roles highlighted in color. (B) The light version of the RDF-based Rosetta Statement metamodel from (A). The statement instance resource indirectly indicates the verb or predicate of the statement, shown in orange. Object arguments (‘rosetta: required object position #’ for resources and ‘rosetta: required literal object position #’ for literals) and adjuncts (‘rosetta: optional object position #’ for resources and ‘rosetta: optional literal object position #’ for literals) can be either object-resources (in blue) or object-literals (in green).
Figure 5 From a formalized natural language statement to the corresponding light version of the RDF Rosetta Statement metamodel. (A) A formalized statement with its syntactic positions and associated semantic roles highlighted in color. (B) The light version of the RDF-based Rosetta Statement metamodel from (A). The statement instance resource indirectly indicates the verb or predicate of the statement, shown in orange. Object arguments (‘rosetta: required object position #’ for resources and ‘rosetta: required literal object position #’ for literals) and adjuncts (‘rosetta: optional object position #’ for resources and ‘rosetta: optional literal object position #’ for literals) can be either object-resources (in blue) or object-literals (in green).CC BY · PubMed Central
Figure 6 Structure of the full version of the Rosetta Statement metamodel. The RDF-based Rosetta Statement schema for the statement from Fig. 5A), according to the full version of the Rosetta Statement metamodel. Compared to the light version (Fig. 5B), it introduces the possibility of having several statement versions by linking each statement version resource to an anchor statement resource via a ‘rosetta: has version’ property. The anchor resource specifies an optional context to which the statement belongs (e.g. a scholarly publication) and identifies through the ‘rosetta: has data schema’ property the Rosetta Statement semantic data schema that it instantiates. Each version has a statement instance to which, indirectly, a number of objects and subjects are linked. Indirect, because each Rosetta Statement class has, depending on the arity of its statement, one or more accompanying object-position classes defined—one for each object argument and adjunct. For a given statement version, the corresponding object-position classes are instantiated and linked to the statement version instance, depending on whether they are arguments (‘rosetta: required object position’) or adjuncts (‘rosetta: optional object position’). The actual object-resources (blue ‘GUPRI’) and object-literals (green ‘literal’) are linked to their respective object-position instance. The same applies to the subject-resource, with the only difference that a general subject position class is used for all Rosetta Statements, independent of their type. This structure supports linking more than one subject resource and more than one object resource or literal to a given subject and object position. Since various metadata can be linked to each statement version resource, including the information that it has been (soft) deleted, the full version of the Rosetta Statement metamodel also supports the versioning of statements and the tracking of the editing history for each object position of each statement in a knowledge graph. Whenever a position is updated, a new version is created in the graph. Metadata associated with the anchor statement resource is not shown.
Figure 6 Structure of the full version of the Rosetta Statement metamodel. The RDF-based Rosetta Statement schema for the statement from Fig. 5A), according to the full version of the Rosetta Statement metamodel. Compared to the light version (Fig. 5B), it introduces the possibility of having several statement versions by linking each statement version resource to an anchor statement resource via a ‘rosetta: has version’ property. The anchor resource specifies an optional context to which the statement belongs (e.g. a scholarly publication) and identifies through the ‘rosetta: has data schema’ property the Rosetta Statement semantic data schema that it instantiates. Each version has a statement instance to which, indirectly, a number of objects and subjects are linked. Indirect, because each Rosetta Statement class has, depending on the arity of its statement, one or more accompanying object-position classes defined—one for each object argument and adjunct. For a given statement version, the corresponding object-position classes are instantiated and linked to the statement version instance, depending on whether they are arguments (‘rosetta: required object position’) or adjuncts (‘rosetta: optional object position’). The actual object-resources (blue ‘GUPRI’) and object-literals (green ‘literal’) are linked to their respective object-position instance. The same applies to the subject-resource, with the only difference that a general subject position class is used for all Rosetta Statements, independent of their type. This structure supports linking more than one subject resource and more than one object resource or literal to a given subject and object position. Since various metadata can be linked to each statement version resource, including the information that it has been (soft) deleted, the full version of the Rosetta Statement metamodel also supports the versioning of statements and the tracking of the editing history for each object position of each statement in a knowledge graph. Whenever a position is updated, a new version is created in the graph. Metadata associated with the anchor statement resource is not shown.CC BY · PubMed Central
Figure 7 Textual and graphical displays of a statement based on its Rosetta Statement schema. Middle: The RDF Rosetta Statement schema of a travels statement (here, applying the light version of the Rosetta Statement approach). Top: A textual display of the travels statement, i.e. a dynamic label that is associated with the Rosetta Statement schema. Note how the subject-position and the object-position slots from the RDF schema align (arrows) with variable-positions (i.e. syntactic positions with associated semantic roles) in the dynamic label template. Bottom: A graphical display of the traveling statement, i.e. a dynamic mind-map pattern that is associated with the Rosetta Statement schema. The alignment of the subject-position and the object-position slots from the RDF Rosetta Statement schema to nodes in the dynamic mind-map pattern is similar to the alignment for the dynamic label template.
Figure 7 Textual and graphical displays of a statement based on its Rosetta Statement schema. Middle: The RDF Rosetta Statement schema of a travels statement (here, applying the light version of the Rosetta Statement approach). Top: A textual display of the travels statement, i.e. a dynamic label that is associated with the Rosetta Statement schema. Note how the subject-position and the object-position slots from the RDF schema align (arrows) with variable-positions (i.e. syntactic positions with associated semantic roles) in the dynamic label template. Bottom: A graphical display of the traveling statement, i.e. a dynamic mind-map pattern that is associated with the Rosetta Statement schema. The alignment of the subject-position and the object-position slots from the RDF Rosetta Statement schema to nodes in the dynamic mind-map pattern is similar to the alignment for the dynamic label template.CC BY · PubMed Central
Figure 8 Schema cross-walks from the light version Rosetta Statement schema for a measurement statement to four other schemata. Middle: The light version Rosetta Statement RDF schema for a measurement statement. Top left: The OBI schema. Top right: Relational database schema. Bottom left: The schema of the Quantity, Units, Dimensions, and Types Ontology (QUDT) (model transferred from [84]). Bottom right: The OBOE schema. Blue arrows indicate the alignment of subject- and object-resource slots and green arrows of object-literal slots. When operationalizing cross-walks, entity mappings are often required to satisfy the constraint specifications of the target schema. For example, if the OBI schema requires a class resource from the Units of Measurement Ontology (UO) for its UNIT slot, the Wikidata resource of the source must be translated into a corresponding UO resource, while for the QUDT schema it would have to be translated into a corresponding QUDT resource. Whereas the relational database model can cover the statement resource of a Rosetta Statement in a dedicated column of the table, to be referencable as a statement, the other three semantic data schemata would have to model the statement in a Named Graph, which would have the statement resource as its GUPRI.
Figure 8 Schema cross-walks from the light version Rosetta Statement schema for a measurement statement to four other schemata. Middle: The light version Rosetta Statement RDF schema for a measurement statement. Top left: The OBI schema. Top right: Relational database schema. Bottom left: The schema of the Quantity, Units, Dimensions, and Types Ontology (QUDT) (model transferred from [84]). Bottom right: The OBOE schema. Blue arrows indicate the alignment of subject- and object-resource slots and green arrows of object-literal slots. When operationalizing cross-walks, entity mappings are often required to satisfy the constraint specifications of the target schema. For example, if the OBI schema requires a class resource from the Units of Measurement Ontology (UO) for its UNIT slot, the Wikidata resource of the source must be translated into a corresponding UO resource, while for the QUDT schema it would have to be translated into a corresponding QUDT resource. Whereas the relational database model can cover the statement resource of a Rosetta Statement in a dedicated column of the table, to be referencable as a statement, the other three semantic data schemata would have to model the statement in a Named Graph, which would have the statement resource as its GUPRI.CC BY · PubMed Central
Figure 9 Input form for specifying a new RDF Rosetta Statement schema and accompanying Rosetta Statement ontology class. Users provide a label (A) and a definition or description (B) for the corresponding Rosetta Statement class, together with some example sentences (C). An overview of the editing progress is provided (D), with initially only the subject position and the verb being specified (here, an advanced editing status is shown). More object positions can be added. Each position in the statement is represented in an item list (E), and items can be expanded to add more information to them (F), specifying their placeholder label, pre- and postposition labels, type restrictions (i.e. input constraints), specification of number of values allowed for the position, and additional text that describes what information is expected for the position (here shown for the Confidence Level position). How the information provided for each position influences the dynamic label textual display is directly shown (D), and users can change the order of the object positions by drag and drop of the items in the list (in the dynamic label (D), required object positions are displayed in darker and optional object positions in lighter color).tor
Figure 9 Input form for specifying a new RDF Rosetta Statement schema and accompanying Rosetta Statement ontology class. Users provide a label (A) and a definition or description (B) for the corresponding Rosetta Statement class, together with some example sentences (C). An overview of the editing progress is provided (D), with initially only the subject position and the verb being specified (here, an advanced editing status is shown). More object positions can be added. Each position in the statement is represented in an item list (E), and items can be expanded to add more information to them (F), specifying their placeholder label, pre- and postposition labels, type restrictions (i.e. input constraints), specification of number of values allowed for the position, and additional text that describes what information is expected for the position (here shown for the Confidence Level position). How the information provided for each position influences the dynamic label textual display is directly shown (D), and users can change the order of the object positions by drag and drop of the items in the list (in the dynamic label (D), required object positions are displayed in darker and optional object positions in lighter color).torCC BY · PubMed Central
Figure 10 Display of RDF Rosetta Statements in the ORKG UI in the view and the edit mode, using the dynamic label. (A) The representation of a weight measurement statement of a particular orange using the measurement statement Rosetta Statement schema, without specifying a confidence level. Empty object-positions and their associated pre- and postposition texts are not displayed in the view mode. (B) The representation of a weight measurement statement of a particular apple, with all positions of the schema being specified. (C) The input form for the measurement statement Rosetta Statement in the ORKG. The subject-position (‘Object’) and each object-position defined in the corresponding Rosetta Statement schema aligns with an input field in the ORKG UI when creating or editing the corresponding Rosetta Statement instance.
Figure 10 Display of RDF Rosetta Statements in the ORKG UI in the view and the edit mode, using the dynamic label. (A) The representation of a weight measurement statement of a particular orange using the measurement statement Rosetta Statement schema, without specifying a confidence level. Empty object-positions and their associated pre- and postposition texts are not displayed in the view mode. (B) The representation of a weight measurement statement of a particular apple, with all positions of the schema being specified. (C) The input form for the measurement statement Rosetta Statement in the ORKG. The subject-position (‘Object’) and each object-position defined in the corresponding Rosetta Statement schema aligns with an input field in the ORKG UI when creating or editing the corresponding Rosetta Statement instance.CC BY · PubMed Central
Figure 11 Step-wise procedure for knowledge graph construction to lower the barrier from the semantic parsing burden. In the first step, the Rosetta Statement approach to knowledge graph construction is applied, using Wikidata as the underlying controlled vocabulary, resulting in a FAIR and CLEAR knowledge graph with limited findability. In a next step, by defining entity mappings between the Wikidata terms used in the knowledge graph and corresponding resources from established OWL ontologies, the FAIR and CLEAR knowledge graph can be transferred, now supporting semantic search. By developing reasoning-supporting semantic data schemata for each Rosetta Statement schema used in the knowledge graph and by defining schema cross-walks between them, the knowledge graph can be transferred into a reasoning-capable FAIR and CLEAR knowledge graph in the final step.
Figure 11 Step-wise procedure for knowledge graph construction to lower the barrier from the semantic parsing burden. In the first step, the Rosetta Statement approach to knowledge graph construction is applied, using Wikidata as the underlying controlled vocabulary, resulting in a FAIR and CLEAR knowledge graph with limited findability. In a next step, by defining entity mappings between the Wikidata terms used in the knowledge graph and corresponding resources from established OWL ontologies, the FAIR and CLEAR knowledge graph can be transferred, now supporting semantic search. By developing reasoning-supporting semantic data schemata for each Rosetta Statement schema used in the knowledge graph and by defining schema cross-walks between them, the knowledge graph can be transferred into a reasoning-capable FAIR and CLEAR knowledge graph in the final step.CC BY · PubMed Central

Verification

Claims the paper makes

0 Valid · 13 Not valid · 69 Not checked · 284 Own results and interpretations

84 more claims

Create a free account to see all 366 claims and to verify the 97 references still open.

See all claims — free account
  • It may state two separate claims as one. 129
  • Probably not a substantial claim. 212

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 81 references

Research works printed but not linked: 5

  1. Smith, 2003, Ontology, 155
  2. Booth, 2019, Session X: easyRDF [internet]
  3. Levin, 1993, English Verb Classes and Alterations: A Preliminary Investigation, 366
  4. Hartig, 2017, Foundations of RDF* and SPARQL*
  5. Mons, 2009, Nano-publication in the e-science era

Cited by

See all 4

Metadata

Authors
Vogt, L., Farfar, K. E., Karanth, P., Konrad, M., Oelen, A., Prinz, M., & Strömert, P.
Affiliations
  • Leibniz Institute for the Analysis of Biodiversity Change (DE)
  • Technische Informationsbibliothek (TIB) (DE)
Corresponding author
Vogt, Lars
Venue
Database
Published
2026
Volume
2026
ISSN
1758-0463 (ISSN-L)
OpenAlex
W4401202450
Cited by
4
References
81
Citations by year
2026: 3
FWCI
3.24
Type
journal article
Language
en
Open access
Locations
  • Database (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion, cc-by)
  • arXiv (Cornell University) (cc-by)
Funders
  • BMBF (13XP5126B)
  • InSuKa (13XP5196F)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Semantic Web and Ontologies
  • Physical Sciences › Computer Science › Information Systems › Research Data Management Practices
  • Physical Sciences › Computer Science › Information Systems › Library Science and Information Systems
Keywords
knowledge graph construction, FAIR data, Open Research Knowledge Graph, semantic search, Wikidata, ontology engineering, large language models
MeSH terms
Biocuration, Semantics, User-Computer Interface, Biological Ontologies

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-18, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-18 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Rosetta Statements: simplifying FAIR knowledge graph construction with a user-centred approach. https://app.cochat.ai/papers/10.1093/database/baag030
The paper
Vogt, L., Farfar, K. E., Karanth, P., Konrad, M., Oelen, A., Prinz, M., & Strömert, P. (2026). Rosetta Statements: simplifying FAIR knowledge graph construction with a user-centred approach. Database. https://doi.org/10.1093/database/baag030

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.