Post-doc position in the MIning MUlti-source and MUlti-modal geo-referenced information (MIMU) project

Innovative geographic knowledge discovery is becoming increasingly possible through the analysis of large-scale data, for instance made available by Earth observation projects leveraging satellites and remote-sensing, provided by ground-level sensors, or given as volunteered geographical information (e.g., geo-referenced multimedia contents posted online on social media platforms). However, problems that involve combining remotely-sensed data with volunteered geographical information are only now starting to be explored, and they still involve a number of practical challenges, e.g. related to appropriate content classification.

Over the recent years, data classification leveraging deep neural networks has also become increasingly popular. These methods have been reported to result in impressive performance gains, when applied to problems related to processing images or natural language. Within GIScience research, deep learning can also have several applications, that are only now starting to be explored (e.g., for improving land coverage analysis, spatial downscaling methods, and general classification tasks that involve combining different types of data).

Within the MIMU project (i.e., acronym for MIning MUlti-source and MUlti-modal geo-referenced information), the researcher that is to be hired will work on the use of machine learning approaches for the discovery and mapping of innovative geographic knowledge through the analysis and processing of large-scale volunteered data (e.g., geo-referenced multimedia contents such as images and textual descriptions, posted on social-media platforms like Flickr, Twitter, or Foursquare), in combination with more traditional sources (e.g., remote- sensing products available in the context of initiatives like ESA’s Sentinel/Copernicus programme, and/or tables with socio-demographic data made available by statistical offices). The complex relations between the different types of information, as well as the temporal and geographical dimensions of the data, introduce new challenges that will explored throughout the project, in an attempt to go beyond the current state-of-the-art.