- Natural Language Processing
- Statistical Machine Translation and Natural Language Processing
- Semantic Web and Ontology Development
- Data Quality Assessment and Improvement
- Graph Neural Network Models and Applications
- Biomedical Ontologies and Text Mining
- QoS-Aware Web Services Composition and Semantic Matching
- Dialogue Act Modeling for Spoken Language Systems
- Web Data Extraction and Crawling Techniques
- Management and Reproducibility of Scientific Workflows
- Visual Question Answering in Images and Videos
- Advances in Transfer Learning and Domain Adaptation
- Chemotherapy-Induced Peripheral Neuropathy in Cancer Treatment
- Data Stream Management Systems and Techniques
- Data Sharing and Stewardship in Science
- Expert Finding in Online Communities
- Recommender System Technologies
- Crisis Communication and Social Media Use
- Artificial Intelligence in Service Industry
- Statistical Mechanics of Complex Networks
- Multi-label Text Classification in Machine Learning
- Collaboration and Dynamics in Wikipedia Research
- Explainable Artificial Intelligence
- Discrete Choice Models in Economics and Health Care
- Ethical Implications of Artificial Intelligence
Universität Hamburg
2021-2024
Leuphana University of Lüneburg
2023-2024
Heinrich Heine University Düsseldorf
2023
Wittenberg University
2023
Walter de Gruyter (Germany)
2023
Translational Research Platform for Veterinary Biologicals (India)
2023
Paderborn University
2017-2021
Fraunhofer Institute for Intelligent Analysis and Information Systems
2020-2021
University of Bonn
2020-2021
Technische Informationsbibliothek (TIB)
2021
Semantic Question Answering (SQA) removes two major access requirements to the Web: mastery of a formal query language like SPARQL and knowledge specific vocabulary. Because complexity natural language, SQA presents difficult challenges many research opportunities. I nstead shared effort, however, essential components are redeveloped, which is an inefficient use researcher’s time resources. This survey analyzes 62 different systems, systematically manually selected using predefined inclusion...
We present GERBIL, an evaluation framework for semantic entity annotation. The rationale behind our is to provide developers, end users and researchers with easy-to-use interfaces that allow the agile, fine-grained uniform of annotation tools on multiple datasets. By these means, we aim ensure both tool developers can derive meaningful insights pertaining extension, integration use applications. In particular, GERBIL provides comparable results so as them easily discover strengths weaknesses...
Modern question answering (QA) systems need to flexibly integrate a number of components specialised fulfil specific tasks in QA pipeline. Key include Named Entity Recognition and Disambiguation, Relation Extraction, Query Building. Since different software exist that implement strategies for each these tasks, it is major challenge select combine the most suitable into system, given characteristics question. We study this optimisation problem train classifiers, which take features as input...
Hyper-relational knowledge graphs (KGs) (e.g., Wikidata) enable associating additional key-value pairs along with the main triple to disambiguate, or restrict validity of a fact. In this work, we propose message passing based graph encoder - StarE capable modeling such hyper-relational KGs. Unlike existing approaches, can encode an arbitrary number information (qualifiers) while keeping semantic roles qualifiers and triples intact. We also demonstrate that benchmarks for evaluating link...
The ability to compare systems from the same domain is of central importance for their introduction into complex applications. In domains named entity recognition and linking, large number orthogonal evaluation w.r.t. measures datasets has led an unclea r landscape regarding abilities weaknesses different approaches. We present Gerbil – improved platform repeatable, storable citable semantic annotation experiments its extension since being release. narrowed this gap by generating concise,...
One of the main tasks when creating and maintaining knowledge bases is to validate facts provide sources for them in order ensure correctness traceability provided knowledge. So far, this task often addressed by human curators a three-step process: issuing appropriate keyword queries statement check using standard search engines, retrieving potentially relevant documents screening those content. The drawbacks process are manifold. Most importantly, it very time-consuming as experts have...
The ability to have the same experience for different user groups (i.e., accessibility) is one of most important characteristics Web-based systems. true Knowledge Graph Question Answering (KGQA) systems that provide access Semantic Web data via natural language interface. While following our research agenda on multilingual aspect accessibility KGQA systems, we identified several ongoing challenges. One them lack benchmarks. In this work, extend popular benchmarks - QALD-9 by introducing...
Over the last decades many machine learning experiments have been published, giving benefit to scientific progress. In order compare machine-learning experiment results with each other and collaborate positively, they need be performed thoroughly on same computing environment, using sample datasets algorithm configurations. Besides this, practical experience shows that scientists engineers tend large output data in their experiments, which is both difficult analyze archive properly without...
Billions of facts pertaining to a multitude domains are now available on the Web as RDF data. However, accessing this data is still difficult endeavour for non-expert users. In order meliorate access data, approaches imposing minimal hurdles their users required. Although many question answering systems over Linked Data have being proposed, retrieving desired significantly challenging. addition, developing and evaluating remains very complex task. To overcome these obstacles, we present...
In this demo, we introduce the DBpedia chatbot, a knowledge-graph-driven chatbot designed to optimize community interaction. The bot was for integration into software facilitate answering of recurrent questions. Four main challenges were addressed when building namely (1) understanding user queries, (2) fetching relevant information based on (3) tailoring responses standards each output platform (i.e. Web, Slack, Facebook) as well (4) developing subsequent interactions with chatbot. With...
The necessity of making the Semantic Web more accessible for lay users, alongside uptake interactive systems and smart assistants Web, have spawned a new generation RDF-based question answering systems. However, fair evaluation these remains challenge due to diffe rent type answers that they provide. Hence, repeating current published experiments or even benchmarking on same datasets complex time-consuming task. We present novel online platform (QA) relies FAIR principles support...
Most question answering (QA) systems over Linked Data, i.e. Knowledge Graphs, approach the task as a conversion from natural language to its corresponding SPARQL query. A common is use query templates generate queries with slots that need be filled. Using instead of running an extensive NLP pipeline or end-to-end model shifts QA problem into classification task, where system needs match input appropriate template. This paper presents automatically learn and classify questions using recursive...
Task-oriented dialogue generation is challenging since the underlying knowledge often dynamic and effectively incorporating into learning process hard. It particularly to generate both human-like informative responses in this setting. Recent research primarily focused on various distillation methods where relationship between facts a base not captured. In paper, we go one step further demonstrate how structural information of graph can improve system’s inference capabilities. Specifically,...
Wikidata is a frequently updated, community-driven, and multilingual knowledge graph. Hence, an attractive basis for Entity Linking, which evident by the recent increase in published papers. This survey focuses on four subjects: (1) Which Linking datasets exist, how widely used are they constructed? (2) Do characteristics of matter design if so, how? (3) How do current approaches exploit specific Wikidata? (4) unexploited existing approaches? reveals that Wikidata-specific not differ their...
Knowledge Graph Question Answering (KGQA) has gained attention from both industry and academia over the past decade. Researchers proposed a substantial amount of benchmarking datasets with different properties, pushing development in this field forward. Many these benchmarks depend on Freebase, DBpedia, or Wikidata. However, KGQA that Freebase DBpedia are gradually less studied used, because is defunct lacks structural validity Therefore, research gravitating toward Wikidata-based...
Entity linking has recently been the subject of a significant body research. Currently, best performing approaches rely on trained mono-lingual models. Porting these to other languages is consequently difficult endeavor as it requires corresponding training data and retraining We address this drawback by presenting novel multilingual, knowledge-based agnostic deterministic approach entity linking, dubbed MAG. MAG based combination context-based retrieval structured knowledge bases graph...
Over the last decades, several billion Web pages have been made available on Web. The ongoing transition from current of unstructured data to Data yet requires scalable and accurate approaches for extraction structured in RDF (Resource Description Framework) these websites. One key steps towards extracting text is disambiguation named entities. We address this issue by presenting AGDISTIS, a novel knowledge-base-agnostic approach entity disambiguation. Our combines Hypertext-Induced Topic...
In this work, we focus on the task of generating SPARQL queries from natural language questions, which can then be executed Knowledge Graphs (KGs). We assume that gold entity and relations have been provided, remaining is to arrange them in right order along with vocabulary, input tokens produce correct query. Pre-trained Language Models (PLMs) not explored depth so far, experiment BART, T5 PGNs (Pointer Generator Networks) BERT embeddings, looking for new baselines PLM era task, DBpedia...