Hamada M. Zahera

Senior Associate Staff and Lead of the NLP Unit, Data Science Group (DICE), Paderborn University

Data Science Group (DICE)
Paderborn University
Paderborn, Germany

👋 Hi! I am a researcher in natural language processing and knowledge graphs. Since March 2024 I lead the NLP unit of the Data Science Group (DICE) at Paderborn University, where I co-supervise PhD researchers and lead the NLP work in funded research projects.

My work connects large language models with knowledge graphs. Current topics include:

question answering and retrieval over knowledge graphs, including text-to-SPARQL
information extraction: entity linking, relation extraction, keyphrase extraction
knowledge editing and reliable, verifiable LLM outputs (including conformal prediction)
multilingual LLMs and ontology learning

I received my PhD (Dr. rer. nat.) in Computer Science from Paderborn University in 2023, supervised by Prof. Axel-Cyrille Ngonga Ngomo and funded by a DAAD doctoral scholarship. Before that, I studied Computer Science at Menoufia University, Egypt.

news

Oct 2026	🎮 New interactive demos for our work on results-aware SPARQL embeddings.
Oct 2026	📄 New preprint: Scaling Multilingual Mixture-of-Experts, on how language coverage changes capacity and compute allocation.
2026	🏆 MARS, multi-hop adaptive retrieval and SPARQL generation for KGQA, accepted at EKAW 2026 and nominated for the best paper award.
2026	🎉 Paper on document-level relation extraction with knowledge graph feedback accepted at ESWC 2026.
2025	🧠 Our benchmark for knowledge editing with logical rules appears at ISWC 2025.
2025	🌍 Two papers at COLING 2025, including LOLA, an open-source massively multilingual LLM.
Mar 2024	🚀 Started leading the NLP unit of the DICE group at Paderborn University.

selected publications

Full list on Google Scholar and DBLP.

2026

arXiv

Scaling Multilingual Mixture-of-Experts: How Language Coverage Changes Capacity and Compute Allocation

J. Dag, N. Srivastava, H. M. Zahera, et al.

arXiv preprint, 2026

arXiv
EKAW

MARS: Multi-hop Adaptive Retrieval and SPARQL Generation for KGQA

N. Srivastava, D. Vollmers, R. Speck, N. Karalis, H. M. Zahera, et al.

EKAW 2026 🏆 Best paper nomination
ESWC

Document-Level Relation Extraction Using Reinforcement Learning with Knowledge Graph Feedback

M. Ali, H. M. Zahera, et al.

ESWC 2026

2025

ISWC

Benchmarking Knowledge Editing using Logical Rules

T. Moteu Ngoli, N. J. Kouagou, H. M. Zahera, et al.

ISWC 2025
COLING

LOLA: An Open-Source Massively Multilingual Large Language Model

N. Srivastava, D. Kuchelev, T. Moteu Ngoli, K. Shetty, M. Röder, H. M. Zahera, et al.

COLING 2025

Code Model
COLING

Contextual Augmentation for Entity Linking using Large Language Models

D. Vollmers, H. M. Zahera, et al.

COLING 2025
ESWC

ANTS: Abstractive Entity Summarization in Knowledge Graphs

A. F. Firmansyah, H. M. Zahera, et al.

ESWC 2025
IEEE Access

Multilingual Relation Extraction: A Survey

M. Ali, R. Speck, H. M. Zahera, et al.

IEEE Access, 2025

2024

EMNLP

Error Analysis of Multilingual Language Models in Machine Translation: A Case Study of English–Amharic

H. M. Alemayehu, H. M. Zahera, A.-C. Ngonga Ngomo

EMNLP 2024
WWW

Universal Knowledge Graph Embeddings

N. J. Kouagou, C. Demir, H. M. Zahera, et al.

Companion Proceedings of the ACM Web Conference (WWW), 2024
EKAW

UniQ-Gen: Unified Query Generation across Multiple Knowledge Graphs

D. Vollmers, H. M. Zahera, et al.

EKAW 2024
SEMANTiCS

Generating SPARQL from Natural Language Using Chain-of-Thoughts Prompting

H. M. Zahera, M. Ali, M. A. Sherif, D. Moussallem, A.-C. Ngonga Ngomo

SEMANTiCS 2024

2022 and earlier

ISWC

MultPAX: Keyphrase Extraction using Language Models and Knowledge Graphs

H. M. Zahera, et al.

ISWC 2022
K-CAP

ASSET: A Semi-supervised Approach for Entity Typing in Knowledge Graphs

H. M. Zahera, S. Heindorf, A.-C. Ngonga Ngomo

K-CAP 2021

demos

🔍 Two SPARQL queries can look almost the same but return different answers, or look very different but return the same answers. That makes plain text similarity a weak signal when you want to find, reuse or cache similar queries.

In Contrastive Learning for Results-Aware SPARQL Embeddings, we run queries from real query logs, label pairs of queries by how much their answers overlap, and fine-tune encoders so that queries with similar answers end up close together. At query time, only the query text is needed. These interactive demos let you explore the idea in your browser:

Results-aware embeddings
See why text similarity is a poor signal for query similarity, and compare a map of real SQuARE test queries under text-based and results-aware embeddings. Impact on retrieval and caching
Look at real queries where text similarity retrieves the wrong query, and see how fine-tuning changes retrieval and result-cache decisions.

Start from the demo overview or browse the code on GitHub.

projects

Research projects

KI-Akademie OWL BMBF, 2024 to 2028
Safe and resource-efficient AI for companies, education and society. My work covers KG question answering, entity linking, knowledge editing and secure retrieval-augmented generation for enterprises.
IKDS BMWE, ZIM cooperation project, 2025 to 2027
A dialogue system that uses a company's own knowledge graph to help optimize machines in operation, built together with industry.
ENEXA EU Horizon Europe, since 2022
Efficient explainable learning on knowledge graphs. I am responsible for the NLP work package.
LOLA and PORQUE 2023 to 2025
An open-source multilingual LLM, and multilingual question answering over knowledge graphs.
DAIKIRI BMBF
Diagnostic AI and embeddings for industrial data.

Open source

SPARQL-Embeddings code
Contrastive embeddings of SPARQL queries by result similarity, combining a CodeBERT text encoder with a GATv2 graph encoder, trained on DBpedia and Wikidata query logs.
LOLA model
Training code for an open multilingual mixture-of-experts LLM covering 160+ languages.
BibCleaner
A Python toolkit (CLI, library and web API) that fixes and enriches BibTeX metadata.
latex-auto-sync
A VS Code extension that keeps LaTeX projects synced with GitHub.

cv

Experience

2024 to now

Senior Associate Staff, Lead of NLP Unit

Data Science Group (DICE), Paderborn University

2020 to 2024

Research Associate

Data Science Group (DICE), Paderborn University

Education

2018 to 2023

PhD (Dr. rer. nat.) in Computer Science, magna cum laude

Paderborn University, Germany. Thesis: Leveraging Social Media in Disaster Situations. DAAD doctoral scholarship.

Service

Area Chair at EACL. Program committee member for ACL, EMNLP, COLING, NeurIPS, AAAI, ISWC and ESWC.