In December 2019 we startet our new research project – HUBBLE – HUman Behaviour-Based explainable machine LEarning. The target of this project is to research and implement new methods and technologies for machine learning based text EXTRACTION from digital sources (websites, social media etc.) and – based on this – a multilingual text CLASSIFICATION to categorize the extracted content. Both developments will be very important new features for MESLIS – our business application for global collecting societies.
HBBO (“human behavior-based optimization”) will be assessed and implemented during the research work to enable the transfer of extracted text features (words and sequences such as word groups, sentences, paragraphs etc., on the basis of which the classification is carried out) from one language to another. To be able to analyze the pipeline in depth (on a algorithm or feature level), elements from “Explainable Machine Learning” (especially: Visual Analytics) will be considered and integrated in the future solution coming up during the project as result of the research work and implementation.
In this project we again cooperate closely with two scientific research instituts: Forensic Science Investigation Lab at the University of Applied Sciences Mittweida and Technical Visualistics at University of Applied Sciences Dresden. The research work will last until 2021 during which we hope to be able to identify interesting and innovative features, publish scientific papers and attend international conferences.
From time to time, we will post some updates here in our blog.