Humboldt-Universität zu Berlin - Mathematisch-Naturwissenschaftliche Fakultät - Institut für Informatik

Verteidigung Masterarbeit: Xing Wang

  • Wann 26.11.2021 von 11:00 bis 23:59
  • Wo online: Zoom
  • iCal

Am 26.11. um 11:00 Uhr (st) wird Xing Wan seine Masterarbeit mit dem Titel

"Distant Supervision for Large-Scale Biomedical Event Extraction and Knowledge Base Population"


Die Verteidigung findet digital per Zoom statt. Eine Zoom-Einladung finden Sie hier. (nur mit Informatik-Account)


Biological pathways can be modeled as large graph structures with molecules represented as nodes linked by biochemical reactions represented as edges. Each year, many new experiments about biological pathways are carried out and their results are gathered in the evergrowing database of biomedical literature. Automated extraction and curation of this body of knowledge requires efficient tools to deal with the large amount of available literature and to be able to correctly parse the reactions and events described in each text.

In this thesis, we propose a novel approach to tackle this task of biomedical event extraction. We develop a multi-turn question answering model which extracts one edge of the event graph at a time. It relies completely on distantly supervised data from PubMed abstracts and PubMedCentral full texts for its training data. Each turn of question answering employs a reader-retriever pipeline where we first filter for relevant literature and secondly examine the found literature for concrete text evidences matching to our current event of interest.