Frank Hopfgartner
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Frank Hopfgartner has a total of 297 co-authors. Here are the top co-authors by number of shared publications:
- Joemon M. Jose
University of Glasgow, Glasgow, United Kingdom
22 shared publications - Cathal Gurrin
Dublin City University, Dublin, Ireland
17 shared publications Amon Rapp
10 shared publicationsAndreas Lommatzsch
Technische Universität Berlin
9 shared publicationsSahin Albayrak
Technische Universität Berlin, Berlin, Germany
9 shared publicationsTill Plumbaum
Technische Universität Berlin, Berlin, Germany
9 shared publications
Latest Publications
- Open Access
- 11 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 596-598
Hidden in Plain Sight: When Hard-to-Notice Prompt Injections Undermine Educational SLM Tutors
- 11 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 596-598
Open AccessHidden in Plain Sight: When Hard-to-Notice Prompt Injections Undermine Educational SLM Tutors
- Open Access
- 2 September 2026
- Published by IOS Press
- p. 87-105
Ontology-Aware Prompting for Knowledge Graph Construction from Text
Constructing Knowledge Graphs (KGs) from unstructured text has traditionally relied on multi-stage pipelines combining Named Entity Recognition, Relation Extraction, and Entity Linking components. These architectures are costly to adapt to new domains and suffer from error propagation across stages. While Large Language Models (LLMs) enable end-to-end triple generation from text, their direct application introduces two key challenges: (1) hallucination, where generated facts are not supported by the input text, and (2) ontology misalignment, where extracted entities and relations violate the constraints of a target schema. To address these limitations, we propose an ontology-aware multi-prompt framework that integrates the target ontology into the prompting process and applies hierarchical verification after triple generation. Specifically, we combine Tree-of-Thoughts-style reasoning for structured exploration, schema-constrained OpenIE, and relaxed triple generation to produce a diverse set of candidate triples while maintaining ontological awareness. A rule-based three-filter filtering module then refines these candidates using cross-prompt agreement, evidence grounding, and surface-form alignment, without relying on external toolchains. Experiments on Text2KGBench, including DBpedia-WebNLG and Wikidata-TekGen, show improvements of up to 2.3 times over LLM-based baselines, together with hallucination rates below 0.20 in most domains and ontology conformance above 0.90 on most DBpedia-WebNLG domains. The framework is released as open-source and available at https://github.com/dice-group/ontology-aware-kg-construction.
- 2 September 2026
- Published by IOS Press
- p. 87-105
Open AccessOntology-Aware Prompting for Knowledge Graph Construction from Text
Constructing Knowledge Graphs (KGs) from unstructured text has traditionally relied on multi-stage pipelines combining Named Entity Recognition, Relation Extraction, and Entity Linking components. These architectures are costly to adapt to new domains and suffer from error propagation across stages. While Large Language Models (LLMs) enable end-to-end triple generation from text, their direct application introduces two key challenges: (1) hallucination, where generated facts are not supported by the input text, and (2) ontology misalignment, where extracted entities and relations violate the constraints of a target schema. To address these limitations, we propose an ontology-aware multi-prompt framework that integrates the target ontology into the prompting process and applies hierarchical verification after triple generation. Specifically, we combine Tree-of-Thoughts-style reasoning for structured exploration, schema-constrained OpenIE, and relaxed triple generation to produce a diverse set of candidate triples while maintaining ontological awareness. A rule-based three-filter filtering module then refines these candidates using cross-prompt agreement, evidence grounding, and surface-form alignment, without relying on external toolchains. Experiments on Text2KGBench, including DBpedia-WebNLG and Wikidata-TekGen, show improvements of up to 2.3 times over LLM-based baselines, together with hallucination rates below 0.20 in most domains and ontology conformance above 0.90 on most DBpedia-WebNLG domains. The framework is released as open-source and available at https://github.com/dice-group/ontology-aware-kg-construction.
- Open Access
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 5328-5331
Trustworthy Personalized Retrieval in the LLM Era: Fairness, Bias Awareness, Privacy, and Multimodality
Personalized search and recommendation are increasingly powered by large language models (LLMs) and multimodal deep learning, yet these advances raise practical concerns about bias, unfair outcomes, privacy risks, and limited transparency . At the same time, personalization pipelines are increasingly end-to-end and adaptive, meaning that small modeling or logging choices can propagate into downstream exposure effects and feedback loops that are hard to detect post hoc. As conversational and generative interfaces become the default access layer, users also face new challenges in judging provenance, uncertainty, and whether recommendations reflect their intent or the model's priors. This tutorial provides a unified, retrieval-centric view of trustworthy personalization: (1) how to measure and mitigate unfairness and bias in ranking, clustering, and classification; (2) how to design bias-aware retrieval experiences that improve user awareness; (3) how to support privacy-aware personal information retrieval (e.g., archived email, lifelogs) without sacrificing utility; and (4) how these concerns evolve for multimodal and retrieval-augmented generative recommendation, including LLM-based retriever/reranker settings. Participants will leave with concrete evaluation protocols, common failure modes, and a practical blueprint for building and auditing end-to-end IR/Rec pipelines for effectiveness, fairness, and privacy.
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 5328-5331
Open AccessTrustworthy Personalized Retrieval in the LLM Era: Fairness, Bias Awareness, Privacy, and Multimodality
Personalized search and recommendation are increasingly powered by large language models (LLMs) and multimodal deep learning, yet these advances raise practical concerns about bias, unfair outcomes, privacy risks, and limited transparency . At the same time, personalization pipelines are increasingly end-to-end and adaptive, meaning that small modeling or logging choices can propagate into downstream exposure effects and feedback loops that are hard to detect post hoc. As conversational and generative interfaces become the default access layer, users also face new challenges in judging provenance, uncertainty, and whether recommendations reflect their intent or the model's priors. This tutorial provides a unified, retrieval-centric view of trustworthy personalization: (1) how to measure and mitigate unfairness and bias in ranking, clustering, and classification; (2) how to design bias-aware retrieval experiences that improve user awareness; (3) how to support privacy-aware personal information retrieval (e.g., archived email, lifelogs) without sacrificing utility; and (4) how these concerns evolve for multimodal and retrieval-augmented generative recommendation, including LLM-based retriever/reranker settings. Participants will leave with concrete evaluation protocols, common failure modes, and a practical blueprint for building and auditing end-to-end IR/Rec pipelines for effectiveness, fairness, and privacy.
- 28 June 2026
- Published by Springer Nature
- p. 27-35
Scaffolding-First Constraint Design for LLM Tutors in Data-Science Problem Solving
- 28 June 2026
- Published by Springer Nature
- p. 27-35
Scaffolding-First Constraint Design for LLM Tutors in Data-Science Problem Solving
- 25 June 2026
- Published by Springer Nature
- p. 190-202
FairFinGAN: Fairness-Aware Synthetic Financial Data Generation
- 25 June 2026
- Published by Springer Nature
- p. 190-202
FairFinGAN: Fairness-Aware Synthetic Financial Data Generation
- 1 June 2026
- Published by IATED Academy in EDULEARN Proceedings
- Vol. 1, 0462
IMPROVING AN LLM PROJECT IDEATION TOOL: EVIDENCE FROM A PILOT STUDY
- 1 June 2026
- Published by IATED Academy in EDULEARN Proceedings
- Vol. 1, 0462
IMPROVING AN LLM PROJECT IDEATION TOOL: EVIDENCE FROM A PILOT STUDY
- Open Access
- 28 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 989-998
Automating Computational Reproducibility in Social Science: Comparing Prompt-Based and Agent-Based Approaches
- 28 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 989-998
Open AccessAutomating Computational Reproducibility in Social Science: Comparing Prompt-Based and Agent-Based Approaches
- 1 March 2026
- Published by IATED Academy in INTED Proceedings
- Vol. 1, 1126
WHEN CLUSTERING FAILS: LESSONS FROM MULTI-ALGORITHM LEARNING ANALYTICS FOR MOOC LEARNING GAP DETECTION
- 1 March 2026
- Published by IATED Academy in INTED Proceedings
- Vol. 1, 1126
WHEN CLUSTERING FAILS: LESSONS FROM MULTI-ALGORITHM LEARNING ANALYTICS FOR MOOC LEARNING GAP DETECTION
- 1 March 2026
- Published by IATED Academy in INTED Proceedings
- Vol. 1, 1134
DOES A MINIMAL LLM ASSISTANT HELP STUDENTS START? A PILOT EVALUATION OF AN LLM-BASED IDEATION TOOL
- 1 March 2026
- Published by IATED Academy in INTED Proceedings
- Vol. 1, 1134
DOES A MINIMAL LLM ASSISTANT HELP STUDENTS START? A PILOT EVALUATION OF AN LLM-BASED IDEATION TOOL
- Open Access
- 14 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 126-131
Bias in Recruitment Systems Utilizing Large Language Models
- 14 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 126-131
Open AccessBias in Recruitment Systems Utilizing Large Language Models