Semantic Rewriting of SPARQL Queries: The Key Role of

Subsumption in Complex Ontology Alignments

Anicet Lepetit Ondo and Laurence Capus

Dep. of Computer and Software Engineering, Laval University, Quebec, G1V 0A6, Quebec, Canada

Keywords: Subsumption Relations, SPARQL Query Rewriting, Aligned Ontology, Large Language Models (LLMs).

Abstract: This article introduces an innovative method for rewriting SPARQL queries in the context of complex

ontology alignment by leveraging hierarchical relations such as subClassOf and subPropertyOf. The method

relies on generalization and specialization links between concepts to retrieve relevant results, even when strict

equivalences are missing. In addition, the use of natural language, assisted by the GPT-4 model, helps address

the syntactic complexity of SPARQL and facilitates interaction with ontologies. Unlike existing approaches

that focus mainly on simple (s: s) or semi-complex (s: c) alignments based on equivalence between source

and target entities, our method reinforces semantic matching by explicitly incorporating subsumption relations.

It also integrates complex (c: c) correspondences, which are often overlooked in the literature, thereby

improving both query coverage and accuracy. Experiments conducted on ontology datasets in the conference

domain confirm the method’s ability to capture a wide range of hierarchical relations. While the method is

designed to be generic, further evaluations on large-scale ontologies are required to assess its robustness and

generalizability.

1 INTRODUCTION

Formal ontologies hold a central place in the

Semantic Web, as they enable the structuring and

formalization of a domain’s semantics in a machine-

interpretable manner. However, the coexistence of

multiple ontologies for the same domain introduces

significant heterogeneity, whether syntactic,

terminological, or conceptual, thus compromising

data sharing and interoperability between systems.

Ontology alignment constitutes a crucial response to

this challenge by establishing correspondences

between entities from different ontologies (Ondo et

al., 2025 ; Amini et al., 2024). These correspondences

may be simple (s: s), semi-complex (s: c), or complex

(c: c), where “s” denotes an elementary entity and “c”

a complex expression, depending on the nature of the

entities being aligned.

Homogeneous querying of aligned complex

ontologies represents a major challenge in

interoperable Semantic Web systems. When a query

is initially formulated on a source ontology, it must

be rewritten to apply to a target ontology, taking into

account syntactic, terminological, and conceptual

heterogeneities between the two. While this rewriting

can be relatively straightforward when based on

elementary correspondences, it becomes significantly

more complex when involving composite entities,

that is, conceptual expressions formed from multiple

interconnected entities (Ondo et al., 2025 ; Thiéblin

et al., 2016). This complexity intensifies when

alignments include hierarchical relations such as

generalization and specialization. These relations are

nevertheless essential to capture the semantic

richness of ontologies, notably by establishing

relevant links even in the absence of strict

equivalences, including when they involve complex

entities. However, they remain largely

underexploited in existing approaches (Ondo et al.,

2025 ; Thiéblin et al., 2021, 2016).

This article presents a method for homogeneous

querying of aligned ontologies, finely exploiting

subsumption relations, including complex

correspondences (c: c). Experimental results

demonstrate that this approach effectively

incorporates various forms of hierarchical relations

between source and target entities, thereby

highlighting the structuring role of subsumptions in

the semantic enrichment of alignments and promoting

interoperability in heterogeneous environments.

The remainder of the article is organized as

follows: we first present related work, followed by the

Ondo, A. L. and Capus, L.

Semantic Rewriting of SPARQL Queries: The Key Role of Subsumption in Complex Ontology Alignments.

DOI: 10.5220/0013676700004000

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 17th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2025) - Volume 2: KEOD and KMIS, pages

99-107

ISBN: 978-989-758-769-6; ISSN: 2184-3228

proposed methodology, experimental results, and

conclude with a discussion and research perspectives.

2 RELATED WORKS

Despite considerable advances over the past decade

in the field of ontology alignments, whether simple

alignments (of type (s: s)) or complex

correspondences (of type (s: c) or (c: c)), research

specifically focused on automatic rewriting

mechanisms for SPARQL queries between a source

ontology and a target ontology remains relatively

scarce.

Indeed, the majority of these works have primarily

focused on identifying and formalizing equivalence

relations between ontological entities, based on

simple (s: s) alignments. In this context, each entity

in the source ontology is simply replaced by its

equivalent in the target ontology by substituting the

original IRI with the one specified in the alignment,

in order to rewrite the query (David et al., 2011).

Some existing works, although they do not

explicitly integrate generalization/specialization

(subsumption) relationships into the rewriting

process, nevertheless exploit complex

correspondences, mainly of the (s: c) type (Thiéblin

et al., 2016, 2017). These approaches use

transformation rules to adapt SPARQL queries,

particularly SELECT queries, to different target

vocabularies. However, a deeper integration of

subsumption relations and complex correspondences

(c: c) may be essential for handling highly structured

ontologies.

Correndo et al. (2010) propose a SPARQL query

rewriting algorithm based on vocabulary alignments

to query heterogeneous RDF sources. While relevant

for data integration, support for complex (c: c)

correspondences and exploration of concept

hierarchies could improve rewriting quality.

Similarly, Fujino et al. (2012) translated queries

between ontologies by focusing on (s: c) and (c: s)

types, without considering complex correspondences

or hierarchical relationships. Thiéblin et al. (2021)

explore two rewriting strategies based on simple

alignments or instance-based approaches, but

highlight the potential of more expressive

correspondences, such as (c: c).

Thakker et al. (2018) also propose a SPARQL

reformulation framework using vocabulary

mappings, though without specifying the types used.

Ondo et al. (2025) propose an innovative

approach for the automatic rewriting of SPARQL

queries across aligned ontologies, taking into account

different types of alignments, both simple (s: s) and

complex. This method places particular emphasis on

complex correspondences (c: c), which constitutes a

notable advancement. However, although

subsumption relations are recognized as relevant in

this context, they are not directly addressed in this

work. Their potential is nevertheless highlighted as a

relevant direction for future research, opening the

way to new avenues of exploration in the field of

semantic interoperability.

This work builds upon the foundation laid in Ondo

et al. (2025), aiming to strengthen the semantic

interoperability of knowledge across aligned

ontologies.

3 ADOPTED METHOD

Our method consists in generating a set 𝑉 and a set

𝑇



, representing respectively the variables and

the triples present in the SPARQL query of the source

ontology 𝑂. Then, taking into account the

correspondences based on subsumption between the

ontologies, we obtain a set 𝑉



and a set 𝑇





corresponding to the rewritten query in the target

ontology 𝑂′.

To formalize this, we define:

𝑇





𝑇



,𝑇



,…,𝑇





(1)

Where T

triplet

is a set of n triples, each triple T

=⟨sᵢ, pᵢ,

oᵢ⟩ for i ∈ {1, 2, …, n}, representing respectively the

subject, predicate, and object in the ontology.

The proposed method combines logical inference,

hierarchical extraction, and semantic analysis to

identify and exploit relevant concepts from two

aligned ontologies, following a structured and concise

six-step process.

3.1 Loading and Inference on the

Ontology

The process starts by importing the aligned ontology

via the Owlready2 library (Lamy, 2020), enabling

OWL model manipulation. The ontology is then

enriched using the Pellet reasoner, which infers

implicit knowledge such as subclass, subproperty,

equivalence, and complex axioms, ensuring full

reasoning for subsequent semantic processing and

query rewriting tasks.

KEOD 2025 - 17th International Conference on Knowledge Engineering and Ontology Development

100

3.2 Hierarchical Identification and

Extraction of Source and Target

Entities

This step involves exploring the ontology to extract

two distinct sets of entities (classes and properties),

differentiated by their membership in the source or

target ontology, as indicated by IRI prefixes. This

separation structures the analysis of correspondences.

For each source entity, one or more target

correspondences are identified based on hierarchical

generalization and specialization relations. When an

equivalence relation connects a complex entity pair,

dedicated functions are applied to derive logical

expressions comprising operators such as

conjunction, disjunction, existential and universal

quantification, and negation. This analytical phase

uncovers explicit complex correspondences, enabling

a fine‑grained and formally interpretable modeling of

the semantic relationships between the aligned

ontologies.

3.3 Building a Cross-Knowledge Base

A correspondence table is generated for each entity of

the source ontology, relying on the subsumption and

equivalence relations established between the two

ontologies. This table contains the source super-

entity, the semantic sub-entities associated with it in

the source ontology, whether atomic elements or

complex expressions, as well as the corresponding or

inferred sub-entities in the target ontology. All of

these correspondences are structured within a formal

knowledge base, intended to be reused in our

SPARQL query rewriting process.

Figure 1 below illustrates an excerpt from the

correspondence table generated by our methodology,

which is based on the transitive propagation of

subsumption relations across multiple levels. This

excerpt highlights all the sub-entities of Event in the

source ontology along with their possible

counterparts (sub-entities) in the target ontology.

More precisely, the entities in the target ontology

identified as subclasses of the Event class in the

source ontology are explicitly defined as subclasses

of the Working_event class in the source ontology due

to the hierarchical structure of subclasses. Since

Working_event is itself a subclass of Event, these

entities are automatically inferred to be subclasses of

Event. This inference relies on the transitivity of

subsumption, which states that:

Let A, B, and C be three sets defA includes the

entities from the target ontology that are identified as

direct subclasses of Working_event. Set B represents

the Working_event entity in the source ontology,

while C corresponds to the Event entity within the

same source ontology.

𝑖𝑓 ∶ ∀𝑥𝐴



𝑥



→𝐵



𝑥



 𝑎𝑛𝑑 ∀𝑥𝐵



𝑥



→𝐶



𝑥





Then, by logical transitivity, ∀𝑥𝐴



𝑥



→𝐶



𝑥





Figure 1: Example of a correspondence table.

3.4 Interaction with the User via

Natural Language and AI Model

When a question is posed, a structured prompt is

automatically generated, incorporating the previously

established semantic correspondence base between

entities of the source and target ontologies. This

prompt is then submitted to the OpenAI’s GPT-4

model, which analyzes the query, identifies the

relevant part of the source graph, and extracts two sets

of corresponding entities (source side and target side),

grouped as a structured paired list. The elements of

these two lists represent the components of a

hierarchy of entities from the source ontology, as well

as their equivalents or subsuming concepts identified

in the target ontology.

3.5 An Automatic SPARQL Query

Generator

Based on the entities extracted respectively from the

source and target ontologies, are automatically

generated SPARQL queries. Each entity is analyzed

to produce, for each scenario type, a query adapted to

the source ontology as well as its translated version

for the target ontology. Structural analysis enables the

extraction of relevant RDF triples, which then serve

as the basis for the dynamic construction of queries.

Thus, the generated SPARQL query adapts to the

Semantic Rewriting of SPARQL Queries: The Key Role of Subsumption in Complex Ontology Alignments

101

semantic and structural characteristics of the analyzed

entities.

3.6 Validation of the Generated

SPARQL Queries

To validate the SPARQL queries automatically

generated from the semantic correspondences

between entities in the source and target ontologies,

our method relied on a rigorous process combining

multiple levels of verification.

We began by analyzing the structure of the data

sources using Protégé 2000. Our method then

automatically generated a correspondence table

linking equivalent source and target subgraphs. This

table was compared against the aligned ontologies to

ensure consistency.

When a user submitted a natural language

question, our method retrieved the relevant source

and target subgraphs from the correspondence table,

followed by a compliance check to ensure logical

soundness.

Validation continued with a structural and visual

comparison of source and target queries, referencing

the subgraph table. This was reinforced by a syntactic

analysis of the automatically generated SPARQL

queries to verify compatibility of the results and

confirm the accuracy of the rewriting. This process

served as a qualitative evaluation of the fidelity of the

generated queries.

4 EXPERIMENT AND RESULTS

This section presents the results obtained by applying

our method to several ontology alignment scenarios,

illustrating its capability to automatically detect,

interpret, and rewrite complex expressions across

aligned ontologies. It begins by introducing the

dataset used in the experiments, followed by a

description of typical cases involving generalization

and specialization. For each scenario, we provide the

corresponding SPARQL query rewritings,

emphasizing their semantic relevance and structural

consistency with respect to the detected alignments.

4.1 The Dataset

In the context of querying complex aligned

ontologies, it is essential to consider the various

correspondence structures that may exist between

them. These configurations constitute a fundamental

basis for developing effective query answering

methods. Consequently, the selection of a suitable

dataset is critical to the success of such experiments.

However, identifying a fully aligned ontology that

enables the evaluation of all possible correspondence

scenarios is often challenging and, in some cases,

infeasible.

To address these challenges, we enriched two

existing datasets (cmt-ConOf.owl) by leveraging a

collection of sixteen ontologies derived from the

well-known conference dataset, commonly used in

research on ontology alignment and querying. This

dataset was prioritized due to its widespread adoption

within the ontology alignment and query rewriting

research community (Thiéblin et al., 2018, 2021 ;

Shvaiko et al., 2023 ; Trojahn et al., 2021), and it

serves as a well-established reference in these

applications.

Furthermore, to assess the robustness and

adaptability of the method across different alignment

configurations, an additional pair of alignments (cmt–

edas.owl), derived from the same dataset but without

enrichment, was also considered.

4.2 Scenarios and Results

Our dataset analysis identified several ontology

alignment scenarios involving subsumption relations.

These include generalization and specialization cases,

where source entities relate to target entities with

partial correspondences, as well as direct

equivalences. These scenarios illustrate the variety

and complexity of relationships present in ontology

alignments.

4.2.1 Scenario 1: Combinations of

SubClassOf Relations with Complex

(c: c) Correspondences

This scenario illustrates a typical case of equivalence

between two classes from aligned ontologies, each

featuring its own internal hierarchy. Defining an

equivalence axiom between the central concepts

enables the propagation of subsumption relations

across both ontologies. Consequently, specializations

defined in one ontology can be inferred in the other,

thus facilitating semantic integration and enabling

coherent query rewriting. Such configurations are

crucial in systems that rely on alignments to query

heterogeneous data sources effectively.

Let us consider the example shown in Figure 2,

inferred from our dataset.

 In the source ontology, the concepts are defined

as follows:

- The class PeerReviewedJournal, denoted as 𝑐



is defined by an equivalence to the following

KEOD 2025 - 17th International Conference on Knowledge Engineering and Ontology Development

102

conjunction: 𝑐



≡𝐽𝑜𝑢𝑟𝑛𝑎𝑙 ⊓

𝑢𝑠𝑒𝑠𝑃𝑟𝑜𝑐𝑒𝑠𝑠. 𝑣𝑎𝑙𝑢𝑒′𝑃𝑒𝑒𝑟𝑅𝑒𝑣𝑖𝑒𝑤′

- The class EngineeringPeerReviewedJournal,

denoted as 𝑐



, is a subclass of 𝑐



:𝑐



⊑𝑐



 In the target ontology:

- The class RefereedJournal, denoted as 𝑐





, is

defined as follows: 𝑐





≡ Publication

⊓

governedBy.some(ReviewPolicy).

- The class MedicalRefereedJournal, denoted as

𝑐





, is a subclass of 𝑐





: 𝑐





⊑𝑐





An ontological alignment is established between the

concepts 𝑐



(in the source ontology) and 𝑐





(in the

target ontology), in the form of an equivalence

axiom:𝑐



≡𝑐





According to description logic, an axiom of the

form EquivalentClasses (A, B) implies that:

 𝐴⊑𝐵 and 𝐵⊑𝐴

 The definitions associated with A and B are

considered interchangeable

 Any entity subsumed by A is also subsumed by

B, and vice versa (Baader et al.,2003).

On this basis, several logical inferences can be drawn:

If 𝑐



≡𝑐





, 𝑐



⊑𝑐



,𝑐





⊑𝑐





, Then, by transitivity and

equivalence:

𝑐





⊑𝑐



; 𝑐



⊑𝑐





; 𝑐



⊑𝑐





; 𝑐





⊑𝑐



After performing a test for the same information need

expressed through different formulations, Figure 3

illustrates the processing within this category

generated by our approach.

Examples of questions used for evaluation

include:

 Which journals make use of the peer-review

process?

 Can you list all journals that are associated with

the peer-review process?

In this category, the rewriting proceeds as follows,

with:



𝑻



𝒙, 𝑪



⋀𝑻



𝒚, 𝑪



x and y are of type C (a class

of the ontology), with 𝐶=𝑐



,𝑐



,…,𝑐



 𝑷𝒙, 𝒑, 𝒗: x has the data property p with the

value v

 𝑹𝒙, 𝒚, 𝒓 : there exists a relation r (object

property) between x and y, with x and y ∈ C

Based on the results illustrated in Figure 3, we

observe that the WHERE clause applied to the source

ontology satisfies the following conditions:

𝑇



=𝑨 ∪𝑩∪𝑪

, with:

For the target ontology, the WHERE clause satisfies

the following conditions

𝑇







𝐷 ∪ 𝐸 ∪ 𝐹



with:

4.2.2 Scenario 2: Combinations of

SubclassOf, SubPropertyOf, and

Complex (c: c) Correspondences

In this category, we apply a rewriting strategy based

on the combination of the subClassOf and

subPropertyOf relations, leveraging the semantic

hierarchies defined between the entities of the source

and target ontologies.

Figure 2: Combinations of subClassOf relations with complex (c: c) correspondences.

𝐴=



𝑥

𝑇𝑥,𝑐



 }

𝐵=



𝑥

𝑇𝑥,𝑐



 }

𝐶=



𝑥

𝑇𝑥,𝑐



∧𝑃𝑥,𝑝,𝑣 }

𝐴=



𝑥

𝑇𝑥, 𝐸𝑛𝑔𝑖𝑛𝑒𝑒𝑟𝑖𝑛𝑔𝑃𝑒𝑒𝑟𝑅𝑒𝑣𝑖𝑒𝑤𝑒𝑑𝐽𝑜𝑢𝑟𝑛𝑎𝑙}

𝐵=



𝑥

𝑇𝑥, 𝑃𝑒𝑒𝑟𝑅𝑒𝑣𝑖𝑒𝑤𝑒𝑑𝐽𝑜𝑢𝑟𝑛𝑎𝑙 }

𝐶=



𝑥

𝑇



𝑥, 𝐽𝑜𝑢𝑟𝑛𝑎𝑙



∧

𝑃𝑥, 𝑢𝑠𝑒𝑠𝑃𝑟𝑜𝑐𝑒𝑠𝑠, 𝑃𝑒𝑒𝑟𝑅𝑒𝑣𝑖𝑒𝑤 }

𝐷=



𝑥

𝑇𝑥,𝑐





 }

𝐸=



𝑥

𝑇𝑥,𝑐





 }

𝐹=



𝑥

𝑇𝑥,𝑐





 ∧ ∃𝑦𝑇𝑦, 𝑐





∧𝑅𝑥,𝑦,𝑟



 }

𝐷=



𝑥

𝑇𝑥, 𝑅𝑒𝑓𝑒𝑟𝑒𝑒𝑑𝐽𝑜𝑢𝑟𝑛𝑎𝑙 }

𝐸=



𝑥

𝑇𝑥, 𝑀𝑒𝑑𝑖𝑐𝑎𝑙𝑅𝑒𝑓𝑒𝑟𝑒𝑒𝑑𝐽𝑜𝑢𝑟𝑛𝑎𝑙 }

𝐹=



𝑥

𝑇𝑥, 𝑃𝑢𝑏𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛 ∧

∃𝑦𝑇𝑦, 𝑅𝑒𝑣𝑖𝑒𝑤𝑃𝑜𝑙𝑖𝑐𝑦 ∧

𝑅𝑥, 𝑦, 𝑔𝑜𝑣𝑒𝑟𝑛𝑒𝑑𝐵𝑦

(3)

(2)

Semantic Rewriting of SPARQL Queries: The Key Role of Subsumption in Complex Ontology Alignments

103

Figure 3: Query rewriting with combinations of subClassOf relations.

Let there be two aligned ontologies, with:

 Source ontology: Let 𝑐



𝑎𝑛𝑑 𝑐



be two classes,

r an object property, and p a data property.

 Target ontology: Let 𝑐





𝑎𝑛𝑑 𝑐





be two classes,

𝒓



an object property, and 𝑝



a data property.

And the following relationships between the entities

of the two ontologies:

 𝑐



≡𝑐





, 𝑐





⊑𝑐



 𝑟



⊑ 𝑟 , with 𝑑𝑜𝑚𝑎𝑖𝑛



𝑟



=𝑐



𝑒𝑡 𝑟𝑎𝑛𝑔𝑒



𝑟



𝑐



; 𝑑𝑜𝑚𝑎𝑖𝑛



𝑟





=𝑐





𝑒𝑡 𝑟𝑎𝑛𝑔𝑒



𝑟





=𝑐





 𝑝



⊑𝑝, 𝑑𝑜𝑚𝑎𝑖𝑛



𝑝



=𝑐



; 𝑑𝑜𝑚𝑎𝑖𝑛𝑝



=

𝑐





Thus, by leveraging subsumption relationships over

data properties, object properties, and classes, our

approach enabled us to infer the following

correspondences:

 For relationships between classes and object

properties, our approach allowed us to identify

two main categories:

- Subsumption with existential quantifiers (∃)



𝑥

𝑇𝑥, 𝑐





 ∧ ∃𝑦𝑇𝑦, 𝑐





∧𝑅𝑥,𝑦,𝑟



 ⊑



𝑥

𝑇𝑥, 𝑐



 ∧ ∃𝑦𝑇𝑦, 𝑐



∧𝑅𝑥,𝑦,𝑟 }

(4)

- Subsumption with universal quantifiers (∀)



𝑥

𝑇𝑥, 𝑐





 ∧ ∀𝑦𝑇𝑦, 𝑐





⇒𝑅𝑥,𝑦,𝑟



 ⊑



𝑥

𝑇𝑥, 𝑐



 ∧ ∀𝑦𝑇𝑦, 𝑐



⇒𝑅𝑥,𝑦,𝑟}

(5)

 For relationships between classes and data

properties, the following category was identified

through our approach:



𝑥

𝑇



𝑥,𝑐







∧𝑃



𝑥,𝑝



,𝑣



⊑



𝑥

𝑇



𝑥,𝑐





∧

𝑃



𝑥,𝑝,𝑣





(6)

These relations show that, if the subsumption

relationships are correctly exploited, a query

formulated on the target ontology can be considered

a valid specialization of an equivalent query on the

source ontology. Figure 4 illustrates the result

deduced from the logic formalized in Equation (4).

The evaluation of this category was based on the

following natural language questions:

 I require the set of individuals of type Person

who are associated with at least one postal

address.

 I would like to obtain the list of people who have

at least one postal address.

4.2.3 Scenario 3: Transitive Propagation of

Subsumptions across Multiple Levels

The idea developed in this category is illustrated in

Figure 1, which presents the complete ontological

hierarchy of the superclass Event, including entities

inferred within the target ontology. Although entities

such as Conference, pc_meeting, and session, which

are specific to the target ontology, are explicitly

defined as direct subclasses of Working_event in the

source ontology, they are inferred as specializations

of Event through transitive propagation of

subsumption relations, in accordance with

ontological inference mechanisms.

Figure 5 illustrates the results generated from the

subsumption relations identified between the source

and target ontologies. In the absence of complex

triples, the results correspond to the union of

instances of atomic classes from both ontologies,

aggregated in the following generalized form.

KEOD 2025 - 17th International Conference on Knowledge Engineering and Ontology Development

104

Figure 4: Rewriting results involving combinations of subClassOf, subPropertyOf.

Figure 5: Rewriting result involving transitive propagation of subsumption relations over multiple levels.

 Source Ontology:

𝑇





𝑥

𝑇



𝑥,𝑐





∨𝑇



𝑥,𝑐





∨𝑇



𝑥,𝑐





∨

…∨𝑇



𝑥,𝑐







(7)

 Target Ontology:

𝑇







𝑥

𝑇



𝑥,𝑐







∨𝑇



𝑥,𝑐







∨𝑇



𝑥,𝑐







∨…∨

𝑇



𝑥,𝑐









(8)

Here, n denotes the number of entities (Classes)

involved in the rewriting process.

4.2.4 Scenario 4: Specific Sublevel

Relying on the logic represented in Figure 1, a query

issued to the entity Working_event automatically

triggers the retrieval of its explicitly defined

subclasses in the target ontology, through ontological

inference mechanisms consistent with the established

subsumption relations. Figure 6 presents the class

Working_Event from the source ontology along with

its subclasses explicitly defined in the target

ontology.

Figure 6:

Rewriting of a Specific Sublevel.

Semantic Rewriting of SPARQL Queries: The Key Role of Subsumption in Complex Ontology Alignments

105

Figure 7: result of the subsumption relations of a specific sublevel.

In this category, the SPARQL query on the source

ontology consists of selecting triples that satisfy the

conditions specified in the WHERE clause, if any

conditions are present. As illustrated in Figure 7, the

WHERE clause of the source query must evaluate a

set of criteria:

𝑇





𝑥

𝑇



𝑥,𝑐





∨𝑇



𝑥,𝑐





∨𝑇



𝑥,𝑐





∨𝑇



𝑥,𝑐







(9)

𝑇





𝑥

𝑇𝑥, 𝑊𝑜𝑟𝑘𝑖𝑛𝑔_𝑒𝑣𝑒𝑛𝑡 ∨

𝑇𝑥, 𝐶𝑜𝑛𝑓𝑒𝑟𝑒𝑛𝑐𝑒 ∨ 𝑇𝑥, 𝑇𝑢𝑡𝑜𝑟𝑖𝑎𝑙 ∨

𝑇



𝑥,𝑊𝑜𝑟𝑘𝑠ℎ𝑜𝑝





While the target query is required to satisfy the

constraints imposed by the target ontology.

𝑇







𝑥

𝑇



𝑥,𝑐







∨𝑇



𝑥,𝑐







∨𝑇



𝑥,𝑐









(10)

𝑇







𝑥

𝑇



𝑥, 𝐶𝑜𝑛𝑓𝑒𝑟𝑒𝑛𝑐𝑒



∨𝑇𝑥,𝑝𝑐





∨𝑇



𝑥, 𝑠𝑒𝑠𝑠𝑖𝑜𝑛





The application of our approach to the example

questions below yields the results shown in Figure 7.

Tested questions were :

 What events are part of the working event group?

 What are the different types of working events?

5 DISCUSSION

As previously mentioned, most existing approaches

focus on simple or partially complex

correspondences, while those addressing complex

correspondences rely primarily on equivalence

relations. By taking into account the majority of

subsumption relations, our method represents a

significant advancement over these previous methods

by broadening the range of relations utilized during

rewriting our method is characterized by strong

modularity, enabling easy adaptation to different

types of ontologies and alignment configurations. It

thus aims to enrich the state of the art, responding to

the limited support for subsumption relations in

automatic rewriting processes, particularly when

involving complex correspondences (c: c) in

SPARQL query rewriting.

In this study, we consider that we have

significantly achieved the objectives we initially set,

by demonstrating the feasibility and effectiveness of

our method to SPARQL query rewriting in the

context of complex ontology alignment involving

subsumption relations. Nevertheless, certain

challenges remain to be addressed, notably the

systematization and full automation of handling these

relations, especially in contexts of very large-scale

aligned ontologies. Another promising avenue for

future research lies in the systematic exploration of

various forms of correspondences involving

cardinality restrictions, particularly those that

leverage minimum, maximum, and exact

cardinalities.

REFERENCES

Amini, R., Norouzi, S.S., Hitzler, P., & Amini, R. (2024).

Towards Complex Ontology Alignment using Large

Language Models. Iberoamerican Conference on

Knowledge Graphs and Semantic Web.

KEOD 2025 - 17th International Conference on Knowledge Engineering and Ontology Development

106

Baader, F., Calvanese, D., McGuinness, D., Nardi, D., et

Patel-Schneider, PF (dir.). (2003). Manuel de logique

de description: théorie, implémentation et applications.

Cambridge University Press.

Correndo, G., Salvadores, M., Millard, I., Glaser, H., &

Shadbolt, N. (2010, March). SPARQL query rewriting

for implementing data integration over linked data.

In Proceedings of the 2010 EDBT/ICDT

Workshops (pp. 1-11).

David, J., Euzenat, J., Scharffe, F., & Trojahn dos Santos,

C. (2011). The alignment API 4.0. Semantic web, 2(1),

3-10.

Fujino, T., & Fukuta, N. (2012, September). A SPARQL

query rewriting approach on heterogeneous ontologies

with mapping reliability. In 2012 IIAI International

Conference on Advanced Applied Informatics (pp. 230-

235). IEEE.

Lamy, J.-B. (2020). Ontologies with Python: Programming

OWL 2.0 Ontologies with Python and Owlready2.

Publisher Apress Berkeley, CA. ISBN 978-1- 4842-

6551-2. https://doi.org/10.1007/978-1-4842-6552-9

Ondo, A.L., Capus, L., & Bousso, M. (2025). Enhancing

SPARQL Query Rewriting for Complex Ontology

Alignments. International journal of Web & Semantic

Technology, Vol.16 (2). DOI: 10.5121/ijwest.

2025.16201

Shvaiko P., M. Abd Nikooie Pour, A. Algergawy, P. Buche,

L. J. Castro, J. Chen, A. Coulet, J. Cu , H. Dong, O.

Fallatah, D. Faria, I. Fundulaki, S. Hertling, Y. He, I.

Horrocks, M. Huschka, L. Ibanescu, S. Jain, E.

Jim´enez-Ruiz, N. Karam, P. Lambrix, H. Li, Y. Li, P.

Monnin, E. Nasr, H. Paulheim, C. Pesquita, T. Saveta,

G. Sousa, C. Trojahn, J. Vatascinova, M. Wu, B.

Yaman, O. Zamazal, L. Zhou,(2023). Results of the

ontology alignment evaluation initiative 2023, in: P.

Shvaiko, J. Euzenat, E. Jim´ enezRuiz, O. Hassanzadeh,

C. Trojahn (Eds.), Proceedings of the 18th International

Workshop on Ontology Matching (OM 2023) co-

located with the 22nd International Semantic Web

Conference (ISWC 2023), Athens, Greece, November

7, 2023, volume 3591 of CEUR Workshop

Proceedings, CEUR-WS.org, 2023, pp. 97–139. URL:

https://ceur-ws.org/Vol-3591/oaei23 paper0.pdf

Thakker D, Schwabe D, Kozaki K, et al. Estimating query

rewriting quality over LOD. Semantic Web.

2018;10(3):529-554. doi:10.3233/SW-180311

Thiéblin, É., Amarger, F., Haemmerlé, O., Hernandez, N.,

& Trojahn, C. (2016). Rewriting SELECT SPARQL

queries from 1: n complex correspondences. CEUR-

WS: Workshop proceedings.

Thiéblin, E., Amarger, F., Hernandez, N., Roussey, C., &

Trojahn Dos Santos, C. (2017). Cross-querying lod

datasets using complex alignments: an application to

agronomic taxa. In Metadata and Semantic Research:

11th International Conference, MTSR 2017, Tallinn,

Estonia, November 28– December 1, 2017,

Proceedings 11 (pp. 25-37). Springer International

Publishing.

Thiéblin, E., Haemmerlé, O., & Trojahn, C. (2021).

Automatic evaluation of complex alignments: An

instance-based approach. Semantic Web, 12(5), 767-

787.

Thiéblin, Haemmerlé, Jane Hernandez, Trojahn dos Santos

(2018). Task-Oriented Complex Ontology Alignment:

Two Alignment Evaluation Sets. European Semantic

Web Conference (ESWC 2018), Jun 2018, Heraklion,

Greece. pp.655–670, ff10.1007/978-3-319-93417-

4_42ff.ffhal-02102383f

Trojahn C., M. Abd Nikooie Pour, A. Algergawy, F.

Amardeilh, R. Amini, O. Fallatah, D. Faria, I.

Fundulaki, I. Harrow, S. Hertling, P. Hitzler, M.

Huschka, L. Ibanescu, E. Jim´ enez-Ruiz, N. Karam, A.

Laadhar, P. Lambrix, H. Li, Y. Li, F. Michel, E. Nasr,

H. Paulheim, C. Pesquita, J. Portisch, C. Roussey, T.

Saveta, P. Shvaiko, A. Splendiani, J. Vatascinov´a, B.

Yaman, O. Zamazal, L. Zhou, (2021). Results of the

ontology alignment evaluation initiative 2021, in: P.

Shvaiko, J. Euzenat, E. Jim´enez-Ruiz, O.

Hassanzadeh, C. Trojahn (Eds.), Proceedings of the

16th International Workshop on Ontology Matching

co-located with the 20th International Semantic Web

Conference (ISWC 2021), Virtual conference, October

25, 2021, volume 3063 of CEUR Workshop

Proceedings, CEUR-WS.org, 2021, pp. 62–108. URL:

http://ceur-ws.org/Vol-3063/oaei21 paper0.pdf.

Semantic Rewriting of SPARQL Queries: The Key Role of Subsumption in Complex Ontology Alignments

107