1 of 32

Knowledge modeling: Thesaurus Nuovo Soggettario and Wikidata, a comparison of two KOSs

Elisabetta Viti

Biblioteca Nazionale Centrale di Firenze

Camillo Carlo Pellizzari di San Girolamo

Scuola Normale Superiore

This session is recorded: Please mute your �microphone and camera when you’re not speaking.

CC BY-SA 4.0 Paolopiccinini 2025 Wikimedia Commons

2 of 32

Summary

Goals and history of the collaboration

Comparison

3 of 32

Goals and history of the collaboration

4 of 32

Goals

Project: semantic and technical interoperability

In particular:

Mapping
Identification of - structural and functional - similarities and differences
Case study: reconciliation between equivalent concepts in the photography domain

Nuovo Soggettario Thesaurus (ThNS)

Wikidata (WD)

5 of 32

A brief history

2005:

2013:

2023:

2025:

Link to Italian Wikipedia (itWP) was added to the tag “References” of the ThNS; then in 2013 itWP started adding reciprocal links to the ThNS.

Property P508 (ThNS identifier) was created in WD. The ThNS IDs, present in itWP articles, were imported into the corresponding WD items.

Tag “Equiv. Wikidata” was created in ThNS with a deeplink to WD items.

Attribution of WD equivalents to ThNS terms and vice versa has become an established practice in the elaboration of ThNS terms.

6 of 32

Institutional agreements and project page

Institutional agreements → more active collaboration

First agreement (“lettera di intenti”) between the Biblioteca Nazionale Centrale di Firenze and Wikimedia Italia in 2013 (renewed in 2014 and 2017)
Second agreement (“convenzione”) in 2025 (pdf)

Project page on Wikidata since 2023: link

7 of 32

ThNS Links to WP and WD

8 of 32

Crosswalks between ThNS and itWP and WD

Statistics as of August 2025:

	from ThNS	from itWP	from WD
to ThNS	/	15,521 (s)	17,475 (s)
to itWP	25,085 (s)	/	all itWP
to WD	13,454 (s)	all itWP	/

9 of 32

Ontologies and thesauri -

comparison

10 of 32

Ontologies

In computer science → formal and explicit specification of a shared knowledge conceptualization.

In the context of data modelling → use of a formal language to set out a formalized representation of a domain of knowledge. They specify:

an abstract representation of the reality of a given domain;
the phenomena (concepts, objects and other entities) are organized in a semantic structure divided into:

classes (domain entities),
instances (actual objects present in the domain),
properties (attributes or qualities),
functions and relationships (which allow entities within the domain to be linked), and
axioms (statements that specify and restrict meanings and the use of classes, instances, functions and relationships)

11 of 32

Thesauri (light ontologies)

“Controlled and structured vocabularies in which concepts are represented by terms, organized so the relationships between concepts are made explicit, and preferred terms are accompanied by lead-in entries for synonyms or quasi-synonyms” (ISO 25964)

Three semantic relationships (hierarchical, associative and equivalence relationships)

12 of 32

Goals comparison between thesauri and ontologies

Thesauri	Ontologies
Vocabulary control	Making assertions about individuals: they are not designed for information retrieval by index terms or class notation but for making assertion about individuals
Mainly to guide the selection of indexing and search terms or the browsing of organized document collections. Becomes machine readable over time	Structured, machine-readable data

13 of 32

Structural comparison between thesauri and ontologies

Thesauri	Ontologies
Concepts expressed by Terms (singular or plurals) and add-ons	Concepts expressed by Classes and Individuals (mainly singular)
Disambiguation of homographic terms	Distinct term labels with different identifiers
Preferred and non-preferred terms	No distinction between preferred and non-preferred terms (assign multiple labels to the same class)
Frequently undifferentiated hierarchical relationships (generic, whole-part and instance)	Hierarchical structure is created through sub-class axioms between classes
Associative relationships	Associative relationships unsuited to an ontology (not semantically precise enough to enable inferencing)

14 of 32

74,490 terms (March 2025)
Updated every six months
Free of charge and licensed under CC BY 4.0
Compliant with ISO 25964 and FAIR principles
Connected to many other thesauri, databases (i.e archives and museums)
Used as a knowledge base for automatic indexing of digital resources

SKOS/RDF (2010- )

(queryable through SPARQL endpoint)

Zthes v. 1.0 (2005- )

MARC21 (2015- )

https://thes.bncf.firenze.sbn.it/

15 of 32

Links to foreign subject indexing tools

Links to museums and archives

Links to thesauri, encyclopedias, classifications, ontologies, etc.

16 of 32

knowledge base
one of the Wikimedia projects
116 million items (March 2025), also including concepts
domain ontologies based mainly on P279 (subclass of)
data licensed under CC0, available in multiple formats and queryable through a SPARQL endpoint

https://www.wikidata.org/wiki/Wikidata:Main_Page

17 of 32

Case study:

Comparison between ThNS and Wikidata in the photography domain

18 of 32

Comparison: work steps (1)

Semi-automatic addition of links to WD items in the ThNS, on the basis of the links to itWP already present in the terms (in tag “Resources”)

through SPARQL queries → WD structured data were extracted and matched with ThNS terms
Process of human verification:

Evaluating exact or partial equivalence in the meaning and form of the concepts/terms/entities
Solving inconsistencies
Searching for semantically more pertinent WD elements

19 of 32

Comparison: work steps (2)

Manual refinement, enlargement and improvement of the ontology of photography in WD, using the ThNS terminology as reference.
Selection of a corpus of 160 terms significant and representative of the conceptual categories involved to represent the discipline in its complexity, in particular regarding:

objects (e.g Cameras);
materials (e.g. Silver salts, Carbon paper);
activities, techniques and related products (e.g. Calotypes and Calotype);
people (e.g. Photographers);
facilities (e.g. Photographic laboratories); and
organisations (e.g. Photographic archives and Photo libraries).

20 of 32

The origins of the corpus

The corpus had different origins:

terms already present in the Thesaurus Nuovo Soggettario
terms suggested by Italian National Bibliography (BNI) and other institutions collaborating on current subject indexing and the ThNS
terms acquired in ThNS which came from the revision of dictionaries and thesauri such as: �- Glossario di terminologia fotografica (dell’AIB)�- Thesaurus per materia-tecnica e materiale composito (dell’ICCD)

21 of 32

Ontology of photography – work done

Analysis, comparison, and mapping between ThNS terms and equivalent items in Wikidata.
Assessing and improving their internal coherence, with changes and interventions:

adding synonyms (UF; aliases)
clarifying definition of concepts (scope and definition notes; descriptions)
reviewing and improvement of hierarchical links (BT; P279)
creation of new Wikidata items; evaluation and possible introduction of new terms in the ThNS
verification/creation of mutual crosswalks
increasing links to other foreign national library thesauri (LCSH, RAMEAU, GND, EMBNE), to specialized thesauri (AAT), to archive and museum databases and to Wikipedia articles

22 of 32

Comparison - structure

Thesaurus Nuovo Soggettario

4 macrocategories and 13 categories as the most general concepts
node labels
mainly monohierarchical

Wikidata

Q35120 (entity) as the most general concept; the first level below has 27 items, the second 524 items, the third 5211 items (as of June 2025)
no node labels
mainly monohierarchical

23 of 32

Comparison - structure

Thesaurus Nuovo Soggettario

mainly concepts expressed by common names (for professions, male and female are differentiated); no proper names (with a few exceptions)
terms can be preferred and non-preferred; each term has an ID

Wikidata

both concepts and proper names
non-preferred terms are aliases in the items, so they don’t have an own ID

24 of 32

Comparison – semantic relations

Thesaurus Nuovo Soggettario

BT (Broader Term), NT (Narrower Term), RT (Related Term) per ISO 25964

Wikidata

BT relations are expressed by the following properties:

P279 (subclass of) – cf. BTG (Broader Term Generic)
P171 (parent taxon) – cf. BTG, but only for taxa
P361 (part of) – cf. BTP (Broader Term Partitive)
P31 (instance of) – cf. BTI (Broader Term Instancial)

NT relations are not explicitly stated
RT relations are expressed by many different properties

25 of 32

Comparison – semantic relations

26 of 32

Comparison – links with other KOSs

Thesaurus Nuovo Soggettario

only univocal links to external databases are possible

not technically possible to link from 1 concept/term of the ThNS to 2(+) concepts/terms of another database
not technically possible to link from 2(+) concepts/terms of the ThNS to 1 concept/term of another database

Wikidata

no hard restriction on links to external databases
checks are possible through constraint violations and SPARQL queries

27 of 32

Comparison – links with other KOSs

Thesaurus Nuovo Soggettario

definition notes, scope notes, notes with reference to other concepts/terms

Wikidata

descriptions (in multiple languages)
some properties can provide further help: e.g. P1889 (different from) has a function similar to the notes with reference to other concepts/terms

28 of 32

Edit statistics

Over the past year:

3.000 changes made in Wikidata by the Research and Semantic Indexing Tools Department of the BNCF
30 new items in Wikidata
530 new cross-walks

29 of 32

Results

experimenting in WD with the creation and refinement of an ontology in a specific domain starting from concepts already organized and structured within a thesaurus;
mutual improvement, development and refinement of structured data;
mutual increase of the links to the databases of Libraries, Archives and Museums; and
laying the foundations for a future re-engineering of the ThNS

30 of 32

Future projects (short-term)

continue the analysis and comparison according to the methodology already tested; the next analyzed topic after photography will be professions
make greater use of the potential of Wikidata in the workflow of preparing the ThNS entries
improve the procedures for identifying and resolving misalignments in the reciprocal links; reflect on the criteria used to construct hierarchies in WD and ThNS, since they are often applied in a dishomogeneous way

31 of 32

Future projects (long-term)

evaluate a possible re-engineering of the ThNS with improvement of the hierarchies and the criteria for assigning BT (divided into their three typologies) in potentially polyhierarchical cases;
improve the structure of the WD ontology based on more rigorous grouping (categories and subcategories) and division principles

32 of 32

Thanks for your attention!

Get in touch with us:

Elisabetta Viti

elisabetta.viti@cultura.gov.it

Camillo Carlo Pellizzari di San Girolamo

camillo.pellizzaridisangirolamo@sns.it

Credits

User:Epìdosis, User:Elisabetta Viti2 CC BY SA 4.0

CC BY-SA 4.0 KommunikationBFH 2020 Wikimedia Commons