As one of seven experts, Z_GIS researcher Sebastian d’Oleire-Oltmanns participated in a bmvit (Austrian Ministry for Transport, Innovation and Technology) initiative at the BeSt Salzburg event. The annual fair addresses pupils on the edge of deciding what to do after high school. Students from the Christian-Doppler High school as well as from the Werkschulheim Felbertal participated in the event in order to learn from and exchange with experts from various disciplines.
The field of UAS-based remote sensing was of high interest for the participating students and generated a broad range of questions and a lively discussion. The exchange provided a detailed view beneath the surface of Geoinformatics to the students.
As an ESRI Development Center, Z_GIS every year has the opportunity to nominate one top student based on his or her software development competence. This year this special honour is awarded to Stefan Zimmer for work leading to his University of Salzburg "Applied Geoinformatics" MSc Thesis "An Efficient Algorithm for Computing Space-Time-Linguistics - Similarities and Labelling Social Media Posts" supervised by Dr Bernd Resch.
Stefan Zimmer has recently completed his master's thesis at Z_GIS. Through his exceptional software development skills, he was able to develop and implement a number of geospatial applications using a variety of software technologies, including the extensive use of Esri technology for spatial data analysis and visualisation.
Spatial distribution of Boston tweets indicating anger/disgust (red), fear (blue), and sadness (yellow)
In his thesis, Stefan developed a machine learning-based algorithm called TwEmLab (Twitter Emotion Labeller) that assigns emotion categories to social media posts like tweets. The workflow comprises a self-designed, multi-dimensional, similarity-based graph creation algorithm and the Junto engine that implements a graph-based labelling algorithm called Label Propagation. In addition to the methodological developments, he was able to convert the complex algorithm in a high-performance implementation that is able to process about 1,000 more datasets than comparable systems in a parallelised algorithm, both for CPU- and GPU-based operations:
Apart from that, Stefan developed a web-based real-time visualisation using Mapbox technology for physiological sensor data and social media posts for use in urban planning and management. Finally, in one of his course projects, he conceived and implemented a location-based game, in which a spaceship can be controlled using a mobile EEG device and voice commands - just using "the power of thought" without manual interaction. The application combines a variety of technologies using Unity, the Emotive SDK, Mapbox, Google Cloud Search API and Open Street Map.