University of Hamburg - to research, to teach, to educate and form, to homepage
Semantic Systems
Department of InformaticsSemantic Systems (SEMS)

Photo: privat/Garrido

Semantic Systems

Theses

We are currently not mentoring any new students.  (updated 2023-09-01)

  • Focus Topics:
    • Eye-Tracking experiments to understand humans writing SPARQL queries over different KGs
    • Eye-Tracking experiment for Explainable AI (XAI): What explains a QA result better? Verbalization, visualization, the SPARQL query, or a  justification (via ChatGPT).  Includes also running a user study.
  • NFDI4DS-related: a PDF annotator with active learning for contents of the Open Research Knowledge Graph (ORKG)
  • NFDI4DS-related: a framework to gather and analyze sets of research documents and transform them in a Research Knowledge Graph
  • NFDI4DS-related: enhancing the NFDI4DS search engine https://github.com/semantic-systems/nfdi-search-engine
  • NFDI4DS-related: link and node prediction for large-scale research knowledge graphs
  • NFDI4DS-related: instance matching to join several large-scale research knowledge graphs
  • Industry-related: Extract relations, i.e., KG triples, from technical manuals
  • Industry-related: Extract entities from technical manuals with a focus on coreference resolution
  • Industry-related: Semantically understanding abbreviations in technical manuals
  • Coypu-related: Extracting knowledge graphs from event texts (event type classification and event argument extraction)
  • Coypu-related: A dataset for reconstructing the Ahrtal catastrophe with open (social) data
  • Coypu-related: A supply chain dataset via NLP/deep learning over a web crawl
  • Coypu-related: Extracting company and product data from the common crawl
  • Question Answering-related: A novel framework for biomedical relation extraction dataset
  • Question Answering-related: Named entity recognition/relation extraction dataset in the biomedical domain to fill a KG for BioKGQA
  • Question Answering-related: Financial Knowledge Graph Question Answering using Reinforcement Learning
  • Question Answering-related: Temporal Knowledge Graph Question Answering using Reinforcement Learning
  • Question Answering-related: Geospatial Knowledge Graph Question Answering using Reinforcement Learning
  • Question Answering-related: Trustworthy Speech Assistants - Analytics of factual QA/chatbot/speech systems and their impact on vulnerable user groups
  • Question Answering-related: Question Answering over Heterogeneous Knowledge Graphs
  • Speech Assistant: to serve as a lab assistant / interactive notebook for natural scientists, e.g. chemists or biologists, coupled to a camera system and Zotero
  • Ethics-related: Investigating a community wiki for mapping data to graphs toward biases
  • Green AI: Ecological impact of Knowledge Extraction and Search Algorithms 
  • MISC: Decentralized, privacy-preserving search engine (SOLID)
  • MISC: Quantum Knowledge Extraction and Question Answering
  • MISC: A digital search engine for individual website visits
  • MISC: Extraction and Completion of Genealogy Graphs from full-texts
  • MISC: Data analytics for letters or the Corona Archive (Cross-disciplinary project with the public history chairs

Reserved, in progress, or processed thesis topics:

  • Survey on relation extraction from short texts [FINISHED 2022]
  • Playing text adventures with reinforcement learning and knowledge graphs [FINISHED 2022]
  • Improving Conversational Question Answering [FINISHED 2022]
  • Creating dialog graphs from stored conversations / Intelligent dialog assistant for form filling over knowledge graphs
  • Geospatial-focused event extraction
  • Online visualization tool for knowledge graphs
  • Ethics-related: Investigating biases in knowledge graph-grounded text annotations
  • Ethics-related: Evaluation of a real-world application regarding dual use-risks
  • Knowledge-infused Language and Speech Models (wav2vec) together with Fraunhofer Center for Maritime Logistics (CML)
  • Entity Extraction via a real-life Data Science product together with eccenca GmbH
  • Extracting knowledge graphs from (event) texts using hybrid AI
  • Fairness in Machine Learning (Eine konzeptionelle Betrachtung von Fairness und ein Vergleich von Supervised Learning Algorithmen)

If you have your own idea, please talk to us!