Declarative Relation Extraction (DeRE, since 2018)
DeRE is a framework in which a task which can be formulated as slot filling is described in an XML dialect. This specification triggers a task specific model to be compiled. This helps transfer of models across tasks and evaluation of approaches in different domains.
Adversarial Satire Detection in German (2019)
We developed a pipeline for training a text classifier which avoids to be confused by confounding variables and tested this on the task of classifying satire under the confounder of the publication source.
Previous work, not under active development
Joint Factorgraph-based Sentiment Analysis (JFSA, 2013-2015)
SCIE, Spinal Cord Injury Information Extraction (2014)
Information extraction system for the recognition and interpretation of spinal cord injury preclinical experiment descriptions in scientific papers. This is the result of a one-year students’ project.
Amazon Review-ID-based crawler (2013)
Factor-Graphs for BioNLP Event Extraction (2011)
KNN: Neural Network Simulator in Delphi(2004)
I developed this neural network platform to learn delphi and to understand the concepts of neural networks. It is completely outdated, but if you like Delphi, it might help somebody to understand the concepts of learning and inference in neural networks. It contains Hopfield nets and feed forward nets, both with graphical user interfaces.