@article{60990, abstract = {{Large Language Models (LLMs) have demonstrated remarkable performance across a wide range of natural language processing tasks. However, their effectiveness in low-resource languages remains underexplored, particularly in complex tasks such as end-to-end Entity Linking (EL), which requires both mention detection and disambiguation against a knowledge base (KB). In earlier work, we introduced IndEL — the first end-to-end EL benchmark dataset for the Indonesian language — covering both a general domain (news) and a specific domain (religious text from the Indonesian translation of the Quran), and evaluated four traditional end-to-end EL systems on this dataset. In this study, we propose ELEVATE-ID, a comprehensive evaluation framework for assessing LLM performance on end-to-end EL in Indonesian. The framework evaluates LLMs under both zero-shot and fine-tuned conditions, using multilingual and Indonesian monolingual models, with Wikidata as the target KB. Our experiments include performance benchmarking, generalization analysis across domains, and systematic error analysis. Results show that GPT-4 and GPT-3.5 achieve the highest accuracy in zero-shot and fine-tuned settings, respectively. However, even fine-tuned GPT-3.5 underperforms compared to DBpedia Spotlight — the weakest of the traditional model baselines — in the general domain. Interestingly, GPT-3.5 outperforms Babelfy in the specific domain. Generalization analysis indicates that fine-tuned GPT-3.5 adapts more effectively to cross-domain and mixed-domain scenarios. Error analysis uncovers persistent challenges that hinder LLM performance: difficulties with non-complete mentions, acronym disambiguation, and full-name recognition in formal contexts. These issues point to limitations in mention boundary detection and contextual grounding. Indonesian-pretrained LLMs, Komodo and Merak, reveal core weaknesses: template leakage and entity hallucination, respectively—underscoring architectural and training limitations in low-resource end-to-end EL.11Code and dataset are available at https://github.com/dice-group/ELEVATE-ID.}}, author = {{Gusmita, Ria Hari and Firmansyah, Asep Fajar and Zahera, Hamada Mohamed Abdelsamee and Ngonga Ngomo, Axel-Cyrille}}, issn = {{0169-023X}}, journal = {{Data & Knowledge Engineering}}, keywords = {{LLMs, Evaluation, End-to-end EL, Indonesian}}, pages = {{102504}}, title = {{{ELEVATE-ID: Extending Large Language Models for End-to-End Entity Linking Evaluation in Indonesian}}}, doi = {{https://doi.org/10.1016/j.datak.2025.102504}}, volume = {{161}}, year = {{2026}}, } @unpublished{62721, abstract = {{We introduce the notion of contrastive ABox explanations to answer questions of the type "Why is a an instance of C, but b is not?". While there are various approaches for explaining positive entailments (why is C(a) entailed by the knowledge base) as well as missing entailments (why is C(b) not entailed) in isolation, contrastive explanations consider both at the same time, which allows them to focus on the relevant commonalities and differences between a and b. We develop an appropriate notion of contrastive explanations for the special case of ABox reasoning with description logic ontologies, and analyze the computational complexity for different variants under different optimality criteria, considering lightweight as well as more expressive description logics. We implemented a first method for computing one variant of contrastive explanations, and evaluated it on generated problems for realistic knowledge bases.}}, author = {{Koopmann, Patrick and Mahmood, Yasir and Ngonga Ngomo, Axel-Cyrille and Tiwari, Balram}}, booktitle = {{Pre-print of paper accepted at AAAI 2026}}, title = {{{Can You Tell the Difference? Contrastive Explanations for ABox Entailments}}}, year = {{2026}}, } @inbook{61325, author = {{Vollmer, Anna-Lisa and Buhl, Heike M. and Alami, Rachid and Främling, Kary and Grimminger, Angela and Booshehri, Meisam and Ngonga Ngomo, Axel-Cyrille}}, booktitle = {{Social Explainable AI}}, editor = {{Rohlfing, Katharina J. and Främling, Kary and Lim, Brian and Alpsancar, Suzana and Thommes, Kirsten}}, pages = {{39--53}}, publisher = {{Springer}}, title = {{{Components of an explanation for co-constructive sXAI}}}, doi = {{10.1007/978-981-96-5290-7_3}}, year = {{2026}}, } @inproceedings{65494, author = {{Manzoor, Ali and Zahera, Hammada M. and Saleem, Muhammad and Mahmood, Yasir and Speck, René and Khan, Hashim and Ngonga Ngomo, Axel-Cyrille}}, booktitle = {{The Semantic Web – 23rd European Semantic Web Conference, ESWC 2026, Dubrovnik , Croatia, May 10-14, 2026, Proceedings}}, keywords = {{dice hamada hashim kiakadamie mahmood manzoor ngonga rene sail saleem}}, title = {{{Document-level Relation Extraction using Reinforcement Learning with Knowledge Graph Feedback}}}, year = {{2026}}, } @inproceedings{65567, abstract = {{We introduce the notion of contrastive ABox explanations to answer questions of the type “Why is a an instance of C, but b is not?”. While there are various approaches for explaining positive entailments (why is C(a) entailed by the knowledge base) as well as missing entailments (why is C(b) not entailed) in isolation, contrastive explanations consider both at the same time, which allows them to focus on the relevant commonalities and differences between a and b. We develop an appropriate notion of contrastive explanations for the special case of ABox reasoning with description logic ontologies, and analyze the computational complexity for different variants under different optimality criteria, considering lightweight as well as more expressive description logics. We implemented a first method for computing one variant of contrastive explanations, and evaluated it on generated problems for realistic knowledge bases.}}, author = {{Koopmann, Patrick and Mahmood, Yasir and Ngonga Ngomo, Axel-Cyrille and Tiwari, Balram}}, booktitle = {{Proceedings of the AAAI Conference on Artificial Intelligence}}, issn = {{2374-3468}}, number = {{23}}, pages = {{19189--19197}}, publisher = {{Association for the Advancement of Artificial Intelligence (AAAI)}}, title = {{{Can You Tell the Difference? Contrastive Explanations for ABox Entailments}}}, doi = {{10.1609/aaai.v40i23.38993}}, volume = {{40}}, year = {{2026}}, } @article{54450, abstract = {{In the last decade, there has been increasing interest in allowing users to understand how the predictions of machine-learned models come about, thus increasing transparency and empowering users to understand and potentially contest those decisions.Dialogue-based approaches, in contrast to traditional one-shot eXplainable Artificial Intelligence (XAI) methods, facilitate interactive, in-depth exploration through multi-turn dialogues, simulating expert conversations. This paper reviews the current state of dialogue-based XAI, presenting a systematic review of 1,339 publications, narrowed down to 14 based on inclusion criteria. We explore theoretical foundations of the systems, propose key dimensions along which different solutions to dialogue-based XAI differ, and identify key use cases, target audiences, system components, and the types of supported queries and responses. Furthermore, we investigate the current paradigms by which systems are evaluated and highlight their key limitations. Key findings include identifying the main use cases, objectives, and audiences targeted by dialogue-based XAI methods, and summarize the main types of questions and information needs. Beyond discussing avenues for future work, we present a meta-architecture for these systems from existing literature and outlined prevalent theoretical frameworks.}}, author = {{Mindlin, Dimitry and Beer, Fabian and Sieger, Leonie Nora and Heindorf, Stefan and Cimiano, Philipp and Esposito, Elena and Ngonga Ngomo, Axel-Cyrille}}, journal = {{Artificial Intelligence Review}}, number = {{3}}, publisher = {{Springer}}, title = {{{Beyond One-Shot Explanations: A Systematic Literature Review of Dialogue-Based XAI Approaches}}}, doi = {{10.1007/s10462-024-11007-7}}, volume = {{58}}, year = {{2025}}, } @inproceedings{59910, abstract = {{The connection between inconsistent databases and Dung’s abstract argumentation framework has recently drawn growing interest. Specifically, an inconsistent database, involving certain types of integrity constraints such as functional and inclusion dependencies, can be viewed as an argumentation framework in Dung’s setting. Nevertheless, no prior work has explored the exact expressive power of Dung’s theory of argumentation when compared to inconsistent databases and integrity constraints. In this paper, we close this gap by arguing that an argumentation framework can also be viewed as an inconsistent database. We first establish a connection between subset-repairs for databases and extensions for AFs considering conflict-free, naive, admissible, and preferred semantics. Further, we define a new family of attribute-based repairs based on the principle of maximal content preservation. The effectiveness of these repairs is then highlighted by connecting them to stable, semi-stable, and stage semantics. Our main contributions include translating an argumentation framework into a database together with integrity constraints. Moreover, this translation can be achieved in polynomial time, which is essential in transferring complexity results between the two formalisms.}}, author = {{Mahmood, Yasir and Hecher, Markus and Ngonga Ngomo, Axel-Cyrille}}, booktitle = {{Proceedings of the AAAI Conference on Artificial Intelligence}}, issn = {{2374-3468}}, number = {{14}}, pages = {{15058--15066}}, publisher = {{Association for the Advancement of Artificial Intelligence (AAAI)}}, title = {{{Dung’s Argumentation Framework: Unveiling the Expressive Power with Inconsistent Databases}}}, doi = {{10.1609/aaai.v39i14.33651}}, volume = {{39}}, year = {{2025}}, } @inproceedings{63027, author = {{Ihtassine, Reda and Firmansyah, Asep Fajar and Srivastava, Nikit and Manzoor, Ali and Ngonga Ngomo, Axel-Cyrille and Sherif, Mohamed Ahmed}}, booktitle = {{Proceedings of the Knowledge Capture Conference 2025}}, publisher = {{ACM}}, title = {{{NL2LS: LLM-based Automatic Linking of Knowledge Graphs}}}, doi = {{10.1145/3731443.3771374}}, year = {{2025}}, } @inproceedings{57324, abstract = {{Generating SPARQL queries is crucial for extracting relevant information from diverse knowledge graphs. However, the structural and semantic differences among these graphs necessitate training or fine-tuning a tailored model for each one. In this paper, we propose UniQ-Gen, a unified query generation approach to generate SPARQL queries across various knowledge graphs. UniQ-Gen integrates entity recognition, disambiguation, and linking through a BERT-NER model and employs cross-encoder ranking to align questions with the Freebase ontology. We conducted several experiments on different benchmark datasets such as LC-QuAD 2.0, GrailQA, and QALD-10. The evaluation results demonstrate that our approach achieves performance equivalent to or better than models fine-tuned for individual knowledge graphs. This finding suggests that fine-tuning a unified model on a heterogeneous dataset of SPARQL queries across different knowledge graphs eliminates the need for separate models for each graph, thereby reducing resource requirements.}}, author = {{Vollmers, Daniel and Srivastava, Nikit and Zahera, Hamada Mohamed Abdelsamee and Moussallem, Diego and Ngonga Ngomo, Axel-Cyrille}}, booktitle = {{Knowledge Engineering and Knowledge Management}}, editor = {{Alam, Mehwish and Rospocher, Marco and van Erp, Marieke and Hollink, Laura and Gesese, Genet Asefa}}, isbn = {{978-3-031-77792-9}}, pages = {{174–189}}, publisher = {{Springer Nature Switzerland}}, title = {{{UniQ-Gen: Unified Query Generation Across Multiple Knowledge Graphs}}}, doi = {{https://doi.org/10.1007/978-3-031-77792-9_11}}, year = {{2025}}, } @unpublished{60718, abstract = {{The ability to generate explanations that are understood by explainees is the quintessence of explainable artificial intelligence. Since understanding depends on the explainee's background and needs, recent research focused on co-constructive explanation dialogues, where an explainer continuously monitors the explainee's understanding and adapts their explanations dynamically. We investigate the ability of large language models (LLMs) to engage as explainers in co-constructive explanation dialogues. In particular, we present a user study in which explainees interact with an LLM in two settings, one of which involves the LLM being instructed to explain a topic co-constructively. We evaluate the explainees' understanding before and after the dialogue, as well as their perception of the LLMs' co-constructive behavior. Our results suggest that LLMs show some co-constructive behaviors, such as asking verification questions, that foster the explainees' engagement and can improve understanding of a topic. However, their ability to effectively monitor the current understanding and scaffold the explanations accordingly remains limited.}}, author = {{Fichtel, Leandra and Spliethöver, Maximilian and Hüllermeier, Eyke and Jimenez, Patricia and Klowait, Nils and Kopp, Stefan and Ngonga Ngomo, Axel-Cyrille and Robrecht, Amelie and Scharlau, Ingrid and Terfloth, Lutz and Vollmer, Anna-Lisa and Wachsmuth, Henning}}, booktitle = {{arXiv:2504.18483}}, pages = {{20}}, title = {{{Investigating Co-Constructive Behavior of Large Language Models in Explanation Dialogues}}}, year = {{2025}}, } @article{61123, abstract = {{Knowledge graphs are used by a growing number of applications to represent structured data. Hence, evaluating the veracity of assertions in knowledge graphs—dubbed fact checking—is currently a challenge of growing importance. However, manual fact checking is commonly impractical due to the sheer size of knowledge graphs. This paper is a systematic survey of recent works on automatic fact checking with a focus on knowledge graphs. We present recent fact-checking approaches, the varied sources they use as background knowledge, and the features they rely upon. Finally, we draw conclusions pertaining to possible future research directions in fact checking knowledge graphs.}}, author = {{Qudus, Umair and Röder, Michael and Saleem, Muhammad and Ngonga Ngomo, Axel-Cyrille}}, issn = {{0360-0300}}, journal = {{ACM Computing Surveys}}, keywords = {{fact checking, knowledge graphs, fact-checkers, check worthiness, evidence retrieval, trust, veracity.}}, publisher = {{Association for Computing Machinery (ACM)}}, title = {{{Fact Checking Knowledge Graphs -- A Survey}}}, doi = {{10.1145/3749838}}, volume = {{58}}, year = {{2025}}, } @inproceedings{59054, author = {{Firmansyah, Asep Fajar and Zahera, Hamada Mohamed Abdelsamee and Sherif, Mohamed and Moussallem, Diego and Ngonga Ngomo, Axel-Cyrille}}, booktitle = {{ESWC2025}}, isbn = {{978-3-031-94575-5}}, keywords = {{firmansyah mousallem ngonga sherif zahera}}, pages = {{133----151}}, publisher = {{pringer Nature Switzerland}}, title = {{{ANTS: Abstractive Entity Summarization in Knowledge Graphs}}}, doi = {{10.1007/978-3-031-94575-5_8}}, year = {{2025}}, } @inproceedings{61202, abstract = {{The number of datasets on the web of data increases continuously. However, the knowledge contained therein cannot be fully utilized without finding links between the entities contained in these datasets. Equivalent entities can not be identified solely by checking the equivalence of IRIs because of the different origins and naming schemes of different data providers. Yet, such equivalences can be discovered by computing the similarity of their attributes. In this paper we propose GLIDE, an approach that links entities from two different datasets by embedding a joint model of these datasets enriched by additional relations describing the similarity of literals. The joint model is embedded into a latent vector space while paying attention to juxtaposing similar literals. We evaluate our approach against state-of-the-art algorithms using real-world datasets commonly used in link discovery literature. The results show that GLIDE outperforms all baselines on 5 of 7 datasets with perfect or near-perfect accuracy. Our approach achieves its best performance on datasets that feature several literals with similarities. Our experiments indicate that researchers should not only pay attention to equal literals in knowledge graph embedding but should also be aware of the distance between similar literals.}}, author = {{Becker, Alexander and Ngonga Ngomo, Axel-Cyrille and Sherif, Mohamed }}, booktitle = {{The Semantic Web – ISWC 2025}}, keywords = {{becker sherif enexa sailproject dice simba ngonga whale}}, title = {{{GLIDE: Knowledge Graph Linking using Distance-Aware Embeddings}}}, year = {{2025}}, } @article{61134, author = {{Manzoor, Ali and Speck, René and Zahera, Hamada Mohamed Abdelsamee and Saleem, Muhammad and Moussallem, Diego and Ngonga Ngomo, Axel-Cyrille}}, issn = {{2169-3536}}, journal = {{IEEE Access}}, pages = {{1--1}}, publisher = {{Institute of Electrical and Electronics Engineers (IEEE)}}, title = {{{Multilingual Relation Extraction - A Survey}}}, doi = {{10.1109/access.2025.3604258}}, year = {{2025}}, } @inproceedings{61234, abstract = {{The ability to generate explanations that are understood by explainees is the quintessence of explainable artificial intelligence. Since understanding depends on the explainee's background and needs, recent research focused on co-constructive explanation dialogues, where an explainer continuously monitors the explainee's understanding and adapts their explanations dynamically. We investigate the ability of large language models (LLMs) to engage as explainers in co-constructive explanation dialogues. In particular, we present a user study in which explainees interact with an LLM in two settings, one of which involves the LLM being instructed to explain a topic co-constructively. We evaluate the explainees' understanding before and after the dialogue, as well as their perception of the LLMs' co-constructive behavior. Our results suggest that LLMs show some co-constructive behaviors, such as asking verification questions, that foster the explainees' engagement and can improve understanding of a topic. However, their ability to effectively monitor the current understanding and scaffold the explanations accordingly remains limited.}}, author = {{Fichtel, Leandra and Spliethöver, Maximilian and Hüllermeier, Eyke and Jimenez, Patricia and Klowait, Nils and Kopp, Stefan and Ngonga Ngomo, Axel-Cyrille and Robrecht, Amelie and Scharlau, Ingrid and Terfloth, Lutz and Vollmer, Anna-Lisa and Wachsmuth, Henning}}, booktitle = {{Proceedings of the 26th Annual Meeting of the Special Interest Group on Discourse and Dialogue}}, publisher = {{Association for Computational Linguistics}}, title = {{{Investigating Co-Constructive Behavior of Large Language Models in Explanation Dialogues}}}, year = {{2025}}, } @article{61819, author = {{Papenkordt, Jörg and Ngonga Ngomo, Axel-Cyrille and Thommes, Kirsten}}, issn = {{0144-929X}}, journal = {{Behaviour & Information Technology}}, pages = {{1--22}}, publisher = {{Informa UK Limited}}, title = {{{Are numerical or verbal explanations of AI the key to appropriate user reliance and error detection?}}}, doi = {{10.1080/0144929x.2025.2568928}}, year = {{2025}}, } @inproceedings{62119, author = {{Ihtassine, Reda and Firmansyah, Asep Fajar and Srivastava, Nikit and Ali, Manzoor and Ngonga Ngomo, Axel-Cyrille and Sherif, Mohamed}}, booktitle = {{Proceedings of the 12th Knowledge Capture Conference 2025, {K-CAP} 2025, The Thirteenth International Conference on Knowledge Capture, December 10 - 12, 2025, Dayton, Ohio, USA}}, keywords = {{Srivastava ali dice enexa firmansyah ihtassine ngonga sailproject sherif whale}}, publisher = {{ACM}}, title = {{{NL2LS: LLM-based Automatic Linking of Knowledge Graphs}}}, year = {{2025}}, } @inproceedings{61041, abstract = {{Large Language Models (LLMs) are increasingly deployed in real-world applications that require access to up-to-date knowledge. However, retraining LLMs is computationally expensive. Therefore, knowledge editing techniques are crucial for maintaining current information and correcting erroneous assertions within pre-trained models. Current benchmarks for knowledge editing primarily focus on recalling edited facts, often neglecting their logical consequences. To address this limitation, we introduce a new benchmark designed to evaluate how knowledge editing methods handle the logical consequences of a single fact edit. Our benchmark extracts relevant logical rules from a knowledge graph for a given edit. Then, it generates multi-hop questions based on these rules to assess the impact on logical consequences. Our findings indicate that while existing knowledge editing approaches can accurately insert direct assertions into LLMs, they frequently fail to inject entailed knowledge. Specifically, experiments with popular methods like ROME and FT reveal a substantial performance gap, up to 24%, between evaluations on directly edited knowledge and on entailed knowledge. This highlights the critical need for semantics-aware evaluation frameworks in knowledge editing.}}, author = {{Moteu Ngoli, Tatiana and Kouagou, N'Dah Jean and Zahera, Hamada Mohamed Abdelsamee and Ngonga Ngomo, Axel-Cyrille}}, booktitle = {{Proceedings of the 24th International Semantic Web Conference (ISWC 2025)}}, isbn = {{978-3-032-09530-5}}, keywords = {{dice sailproject moteu kouagou zahera ngonga}}, location = {{Nara, Japan}}, pages = {{pp 41--56}}, publisher = {{Springer, Cham}}, title = {{{Benchmarking Knowledge Editing using Logical Rules}}}, doi = {{https://doi.org/10.1007/978-3-032-09530-5_3}}, year = {{2025}}, } @inproceedings{61753, abstract = {{This paper presents LOLA, a massively multilingual large language model trained on more than 160 languages using a sparse Mixture-of-Experts Transformer architecture. Our architectural and implementation choices address the challenge of harnessing linguistic diversity while maintaining efficiency and avoiding the common pitfalls of multilinguality. Our analysis of the evaluation results shows competitive performance in natural language generation and understanding tasks. Additionally, we demonstrate how the learned expert-routing mechanism exploits implicit phylogenetic linguistic patterns to potentially alleviate the curse of multilinguality. We provide an in-depth look at the training process, an analysis of the datasets, and a balanced exploration of the model{’}s strengths and limitations. As an open-source model, LOLA promotes reproducibility and serves as a robust foundation for future research. Our findings enable the development of compute-efficient multilingual models with strong, scalable performance across languages.}}, author = {{Srivastava, Nikit and Kuchelev, Denis and Moteu Ngoli, Tatiana and Shetty, Kshitij and Röder, Michael and Zahera, Hamada Mohamed Abdelsamee and Moussallem, Diego and Ngonga Ngomo, Axel-Cyrille}}, booktitle = {{Proceedings of the 31st International Conference on Computational Linguistics}}, editor = {{Rambow, Owen and Wanner, Leo and Apidianaki, Marianna and Al-Khalifa, Hend and Eugenio, Barbara Di and Schockaert, Steven}}, pages = {{6420–6446}}, publisher = {{Association for Computational Linguistics}}, title = {{{LOLA – An Open-Source Massively Multilingual Large Language Model}}}, year = {{2025}}, } @inbook{63507, author = {{Pandit, Gaurav and Röder, Michael and Ngonga Ngomo, Axel-Cyrille}}, booktitle = {{Lecture Notes in Computer Science}}, isbn = {{9783031945748}}, issn = {{0302-9743}}, publisher = {{Springer Nature Switzerland}}, title = {{{Evaluating Approximate Nearest Neighbour Search Systems on Knowledge Graph Embeddings}}}, doi = {{10.1007/978-3-031-94575-5_4}}, year = {{2025}}, }