Danai Symeonidou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Danai Symeonidou has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- Bill Gates Happi Happi
DIADE, IRD, CIRAD, University of Montpellier, France
4 shared publications - Pierre Larmande
DIADE, IRD, CIRAD, Univ. Montpellier, Ave Agropolis, Montpellier, France.
4 shared publications - Geraud Fokou Pelap
URIFIA, Department of Computer Science, University of Dschang , Dschang 00237
3 shared publications Bill Gates
2 shared publications- Clement Jonquet
LIRMM University of Montpellier CNRS France
2 shared publications - Raphaël Conde Salazar
MISTEA, University of Montpellier, INRAE & Institut Agro, France, 2, place Pierre Viala, 34060 Montpellier Cedex 2, France
2 shared publications
Latest Publications
- 31 March 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Automatic key discovery for Data Linking
Découverte des clés pour le Liage de Données: Dans les dernières années, le Web de données a connu une croissance fulgurante arrivant à un grand nombre des triples RDF. Un des objectifs les plus importants des applications RDF est l’intégration de données décrites dans les différents jeux de données RDF et la création des liens sémantiques entre eux. Ces liens expriment des correspondances sémantiques entre les entités d’ontologies ou entre les données. Parmi les différents types de liens sémantiques qui peuvent être établis, les liens d’identité expriment le fait que différentes ressources réfèrent au même objet du monde réel. Le nombre de liens d’identité déclaré reste souvent faible si on le compare au volume des données disponibles. Plusieurs approches de liage de données déduisent des liens d’identité en utilisant des clés. Une clé représente un ensemble de propriétés qui identifie de façon unique chaque ressource décrite par les données. Néanmoins, dans la plupart des jeux de données publiés sur le Web, les clés ne sont pas disponibles et leur déclaration peut être difficile, même pour un expert.L’objectif de cette thèse est d’étudier le problème de la découverte automatique de clés dans des sources de données RDF et de proposer de nouvelles approches efficaces pour résoudre ce problème. Les données publiées sur le Web sont général volumineuses, incomplètes, et peuvent contenir des informations erronées ou des doublons. Aussi, nous nous sommes focalisés sur la définition d’approches capables de découvrir des clés dans de tels jeux de données. Par conséquent, nous nous focalisons sur le développement d’approches de découverte de clés capables de gérer des jeux de données contenant des informations nombreuses, incomplètes ou erronées. Notre objectif est de découvrir autant de clés que possible, même celles qui sont valides uniquement dans des sous-ensembles de données.Nous introduisons tout d’abord KD2R, une approche qui permet la découverte automatique de clés composites dans des jeux de données RDF pour lesquels l’hypothèse du nom Unique est respectée. Ces données peuvent être conformées à des ontologies différentes. Pour faire face à l’incomplétude des données, KD2R propose deux heuristiques qui per- mettent de faire des hypothèses différentes sur les informations éventuellement absentes. Cependant, cette approche est difficilement applicable pour des sources de données de grande taille. Aussi, nous avons développé une seconde approche, SAKey, qui exploite différentes techniques de filtrage et d’élagage. De plus, SAKey permet à l’utilisateur de découvrir des clés dans des jeux de données qui contiennent des données erronées ou des doublons. Plus précisément, SAKey découvre des clés, appelées "almost keys", pour lesquelles un nombre d’exceptions est toléré.
- 31 March 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Automatic key discovery for Data Linking
Découverte des clés pour le Liage de Données: Dans les dernières années, le Web de données a connu une croissance fulgurante arrivant à un grand nombre des triples RDF. Un des objectifs les plus importants des applications RDF est l’intégration de données décrites dans les différents jeux de données RDF et la création des liens sémantiques entre eux. Ces liens expriment des correspondances sémantiques entre les entités d’ontologies ou entre les données. Parmi les différents types de liens sémantiques qui peuvent être établis, les liens d’identité expriment le fait que différentes ressources réfèrent au même objet du monde réel. Le nombre de liens d’identité déclaré reste souvent faible si on le compare au volume des données disponibles. Plusieurs approches de liage de données déduisent des liens d’identité en utilisant des clés. Une clé représente un ensemble de propriétés qui identifie de façon unique chaque ressource décrite par les données. Néanmoins, dans la plupart des jeux de données publiés sur le Web, les clés ne sont pas disponibles et leur déclaration peut être difficile, même pour un expert.L’objectif de cette thèse est d’étudier le problème de la découverte automatique de clés dans des sources de données RDF et de proposer de nouvelles approches efficaces pour résoudre ce problème. Les données publiées sur le Web sont général volumineuses, incomplètes, et peuvent contenir des informations erronées ou des doublons. Aussi, nous nous sommes focalisés sur la définition d’approches capables de découvrir des clés dans de tels jeux de données. Par conséquent, nous nous focalisons sur le développement d’approches de découverte de clés capables de gérer des jeux de données contenant des informations nombreuses, incomplètes ou erronées. Notre objectif est de découvrir autant de clés que possible, même celles qui sont valides uniquement dans des sous-ensembles de données.Nous introduisons tout d’abord KD2R, une approche qui permet la découverte automatique de clés composites dans des jeux de données RDF pour lesquels l’hypothèse du nom Unique est respectée. Ces données peuvent être conformées à des ontologies différentes. Pour faire face à l’incomplétude des données, KD2R propose deux heuristiques qui per- mettent de faire des hypothèses différentes sur les informations éventuellement absentes. Cependant, cette approche est difficilement applicable pour des sources de données de grande taille. Aussi, nous avons développé une seconde approche, SAKey, qui exploite différentes techniques de filtrage et d’élagage. De plus, SAKey permet à l’utilisateur de découvrir des clés dans des jeux de données qui contiennent des données erronées ou des doublons. Plus précisément, SAKey découvre des clés, appelées "almost keys", pour lesquelles un nombre d’exceptions est toléré.
- Open Access
- 26 December 2024
- Published by Oxford University Press (OUP) in Bioinformatics Advances
- Vol. 5 (1) , vbaf096
GRU-SCANET: unleashing the power of GRU-based sinusoidal capture network for precision-driven named entity recognition
Motivation: Pre-trained Language Models (PLMs) have achieved remarkable performance across various natural language processing tasks. However, they encounter challenges in biomedical named entity recognition (NER), such as high computational costs and the need for complex fine-tuning. These limitations hinder the efficient recognition of biological entities, especially within specialized corpora. To address these issues, we introduce GRU-SCANET (Gated Recurrent Unit-based Sinusoidal Capture Network), a novel architecture that directly models the relationship between input tokens and entity classes. Our approach offers a computationally efficient alternative for extracting biological entities by capturing contextual dependencies within biomedical texts. Results: GRU-SCANET combines positional encoding, bidirectional GRUs (BiGRUs), an attention-based encoder, and a conditional random field (CRF) decoder to achieve high precision in entity labeling. This design effectively mitigates the challenges posed by unbalanced data across multiple corpora. Our model consistently outperforms leading benchmarks, achieving better performance than BioBERT (8/8 evaluations), PubMedBERT (5/5 evaluations), and the previous state-of-the-art (SOTA) models (8/8 evaluations), including Bern2 (5/5 evaluations). These results highlight the strength of our approach in capturing token-entity relationships more effectively than existing methods, advancing the state of biomedicalNER. Availability and implementation: https://github.com/ANR-DIG-AI/GRU-SCANET.
- 26 December 2024
- Published by Oxford University Press (OUP) in Bioinformatics Advances
- Vol. 5 (1) , vbaf096
Open AccessGRU-SCANET: unleashing the power of GRU-based sinusoidal capture network for precision-driven named entity recognition
Motivation: Pre-trained Language Models (PLMs) have achieved remarkable performance across various natural language processing tasks. However, they encounter challenges in biomedical named entity recognition (NER), such as high computational costs and the need for complex fine-tuning. These limitations hinder the efficient recognition of biological entities, especially within specialized corpora. To address these issues, we introduce GRU-SCANET (Gated Recurrent Unit-based Sinusoidal Capture Network), a novel architecture that directly models the relationship between input tokens and entity classes. Our approach offers a computationally efficient alternative for extracting biological entities by capturing contextual dependencies within biomedical texts. Results: GRU-SCANET combines positional encoding, bidirectional GRUs (BiGRUs), an attention-based encoder, and a conditional random field (CRF) decoder to achieve high precision in entity labeling. This design effectively mitigates the challenges posed by unbalanced data across multiple corpora. Our model consistently outperforms leading benchmarks, achieving better performance than BioBERT (8/8 evaluations), PubMedBERT (5/5 evaluations), and the previous state-of-the-art (SOTA) models (8/8 evaluations), including Bern2 (5/5 evaluations). These results highlight the strength of our approach in capturing token-entity relationships more effectively than existing methods, advancing the state of biomedicalNER. Availability and implementation: https://github.com/ANR-DIG-AI/GRU-SCANET.
- Open Access
GRU-SCANET: Unleashing the Power of GRU-based Sinusoidal CApture Network for Precision-driven Named Entity Recognition
Motivation: Pre-trained Language Models (PLMs) have achieved remarkable performance across various natural language processing tasks. However, they encounter challenges in biomedical Named Entity Recognition (NER), such as high computational costs and the need for complex fine-tuning. These limitations hinder the efficient recognition of biological entities, especially within specialized corpora. To address these issues, we introduce GRU-SCANET (Gated Recurrent Unit-based Sinusoidal Capture Network), a novel architecture that directly models the relationship between input tokens and entity classes. Our approach offers a computationally efficient alternative for extracting biological entities by capturing contextual dependencies within biomedical texts. Results: GRU-SCANET combines positional encoding, bidirectional GRUs (BiGRUs), an attention-based encoder, and a conditional random field (CRF) decoder to achieve high precision in entity labeling. This design effectively mitigates the challenges posed by unbalanced data across multiple corpora. Our model consistently outperforms leading benchmarks, achieving better performance than BioBERT (8/8 evaluations), PubMedBERT (5/5 evaluations), and the previous state-of-the-art (SOTA) models (8/8 evaluations), including Bern2 (5/5 evaluations). These results highlight the strength of our approach in capturing token-entity relationships more effectively than existing methods, advancing the state of biomedical NER.
Open AccessGRU-SCANET: Unleashing the Power of GRU-based Sinusoidal CApture Network for Precision-driven Named Entity Recognition
Motivation: Pre-trained Language Models (PLMs) have achieved remarkable performance across various natural language processing tasks. However, they encounter challenges in biomedical Named Entity Recognition (NER), such as high computational costs and the need for complex fine-tuning. These limitations hinder the efficient recognition of biological entities, especially within specialized corpora. To address these issues, we introduce GRU-SCANET (Gated Recurrent Unit-based Sinusoidal Capture Network), a novel architecture that directly models the relationship between input tokens and entity classes. Our approach offers a computationally efficient alternative for extracting biological entities by capturing contextual dependencies within biomedical texts. Results: GRU-SCANET combines positional encoding, bidirectional GRUs (BiGRUs), an attention-based encoder, and a conditional random field (CRF) decoder to achieve high precision in entity labeling. This design effectively mitigates the challenges posed by unbalanced data across multiple corpora. Our model consistently outperforms leading benchmarks, achieving better performance than BioBERT (8/8 evaluations), PubMedBERT (5/5 evaluations), and the previous state-of-the-art (SOTA) models (8/8 evaluations), including Bern2 (5/5 evaluations). These results highlight the strength of our approach in capturing token-entity relationships more effectively than existing methods, advancing the state of biomedical NER.
- 27 November 2024
- Published by Springer Nature
- p. 271-288
Leveraging Knowledge Graphs for Earth System Dataset Discovery
- 27 November 2024
- Published by Springer Nature
- p. 271-288
Leveraging Knowledge Graphs for Earth System Dataset Discovery
- Open Access
GRU-SCANET: Unleashing the Power of GRU-based Sinusoidal CApture Network for Precision-driven Named Entity Recognition
Motivation: Pre-trained Language Models (PLMs) have achieved remarkable performance across various natural language processing tasks. However, they encounter challenges in biomedical Named Entity Recognition (NER), such as high computational costs and the need for complex fine-tuning. These limitations hinder the efficient recognition of biological entities, especially within specialized corpora. To address these issues, we introduce GRU-SCANET (Gated Recurrent Unit-based Sinusoidal Capture Network), a novel architecture that directly models the relationship between input tokens and entity classes. Our approach offers a computationally efficient alternative for extracting biological entities by capturing contextual dependencies within biomedical texts. Results: GRU-SCANET combines positional encoding, bidirectional GRUs (BiGRUs), an attention-based encoder, and a conditional random field (CRF) decoder to achieve high precision in entity labeling. This design effectively mitigates the challenges posed by unbalanced data across multiple corpora. Our model consistently outperforms leading benchmarks, achieving better performance than BioBERT (8/8 evaluations), PubMedBERT (5/5 evaluations), and the previous stateof-the-art (SOTA) models (8/8 evaluations), including Bern2 (5/5 evaluations). These results highlight the strength of our approach in capturing token-entity relationships more effectively than existing methods, advancing the state of biomedical NER.
Open AccessGRU-SCANET: Unleashing the Power of GRU-based Sinusoidal CApture Network for Precision-driven Named Entity Recognition
Motivation: Pre-trained Language Models (PLMs) have achieved remarkable performance across various natural language processing tasks. However, they encounter challenges in biomedical Named Entity Recognition (NER), such as high computational costs and the need for complex fine-tuning. These limitations hinder the efficient recognition of biological entities, especially within specialized corpora. To address these issues, we introduce GRU-SCANET (Gated Recurrent Unit-based Sinusoidal Capture Network), a novel architecture that directly models the relationship between input tokens and entity classes. Our approach offers a computationally efficient alternative for extracting biological entities by capturing contextual dependencies within biomedical texts. Results: GRU-SCANET combines positional encoding, bidirectional GRUs (BiGRUs), an attention-based encoder, and a conditional random field (CRF) decoder to achieve high precision in entity labeling. This design effectively mitigates the challenges posed by unbalanced data across multiple corpora. Our model consistently outperforms leading benchmarks, achieving better performance than BioBERT (8/8 evaluations), PubMedBERT (5/5 evaluations), and the previous stateof-the-art (SOTA) models (8/8 evaluations), including Bern2 (5/5 evaluations). These results highlight the strength of our approach in capturing token-entity relationships more effectively than existing methods, advancing the state of biomedical NER.
- Open Access
- 10 October 2024
- Published by Elsevier in Data in Brief
- Vol. 57, 111017
RDF graph pair profile dataset for the data linking community
- 10 October 2024
- Published by Elsevier in Data in Brief
- Vol. 57, 111017
Open AccessRDF graph pair profile dataset for the data linking community
- Open Access
Fine-tuning Large Language Models (LLMs) for Entity Alignment in Knowledge Graphs (KG)
Searching for similar entities between diverse and heterogeneous data sources in knowledge graphs (KGs) has always been a major challenge. The advent of large language models (LLMs) has opened up new avenues for research and improvement of existing solutions. For example, fine-tuning an LLM was rapidly adopted due to its ability to specialize in specific tasks. This challenge notably involves capturing subtle similarities between the entities concerned, at the linguistic, syntactic, and semantic levels. In this article, we propose an approach to fine-tune GPT-2 and BERT language models to tackle the generalization of entity alignment problems on many different datasets by one single model. We also propose a protocol for entity alignment based on the Kolmogorov Arnold Network (KAN) to address the limitation of interpretability, redundancy, and computational cost of LLMs. The evaluations conducted in this study highlight the significant contribution of the fine-tuned GPT-2 language model, ensuring superior performance on entity alignment challenges compared to BERT and KAN.
Open AccessFine-tuning Large Language Models (LLMs) for Entity Alignment in Knowledge Graphs (KG)
Searching for similar entities between diverse and heterogeneous data sources in knowledge graphs (KGs) has always been a major challenge. The advent of large language models (LLMs) has opened up new avenues for research and improvement of existing solutions. For example, fine-tuning an LLM was rapidly adopted due to its ability to specialize in specific tasks. This challenge notably involves capturing subtle similarities between the entities concerned, at the linguistic, syntactic, and semantic levels. In this article, we propose an approach to fine-tune GPT-2 and BERT language models to tackle the generalization of entity alignment problems on many different datasets by one single model. We also propose a protocol for entity alignment based on the Kolmogorov Arnold Network (KAN) to address the limitation of interpretability, redundancy, and computational cost of LLMs. The evaluations conducted in this study highlight the significant contribution of the fine-tuned GPT-2 language model, ensuring superior performance on entity alignment challenges compared to BERT and KAN.
- Open Access
- 11 September 2023
- Published by IOS Press
- p. 194-209
Classification of Linking Problem Types for Linking Semantic Data
As the number of RDF datasets published on the semantic web continues to grow, it becomes increasingly important to efficiently link similar entities between these datasets. However, the performance of existing data linking tools, often developed for general purposes, seems to have reached a plateau, suggesting the need for more modular and efficient solutions. In this paper, we propose –and formalize in OWL– a classification of the different Linking Problem Types (LPTs) to help the linked data community identify upstream the problems and develop more efficient solutions. Our classification is based on the description of heterogeneity reported in the literature –especially five articles– and identifies five main types of linking problems: predicate value problems, predicate problems, class problems, subgraph problems, and graph problems. By classifying LPTs, we provide a framework for understanding and addressing the challenges associated with semantic data linking. It can be used to develop new solutions based on existing modularized tools addressing specific LPTs, thus improving the overall efficiency of data linking.
- 11 September 2023
- Published by IOS Press
- p. 194-209
Open AccessClassification of Linking Problem Types for Linking Semantic Data
As the number of RDF datasets published on the semantic web continues to grow, it becomes increasingly important to efficiently link similar entities between these datasets. However, the performance of existing data linking tools, often developed for general purposes, seems to have reached a plateau, suggesting the need for more modular and efficient solutions. In this paper, we propose –and formalize in OWL– a classification of the different Linking Problem Types (LPTs) to help the linked data community identify upstream the problems and develop more efficient solutions. Our classification is based on the description of heterogeneity reported in the literature –especially five articles– and identifies five main types of linking problems: predicate value problems, predicate problems, class problems, subgraph problems, and graph problems. By classifying LPTs, we provide a framework for understanding and addressing the challenges associated with semantic data linking. It can be used to develop new solutions based on existing modularized tools addressing specific LPTs, thus improving the overall efficiency of data linking.
- 1 April 2022
- Published by Springer Nature
- p. 57-64
An Ontology to Structure Biological Data: The Contribution of Mathematical Models
- 1 April 2022
- Published by Springer Nature
- p. 57-64
An Ontology to Structure Biological Data: The Contribution of Mathematical Models