Skip to main content
European Commission logo print header

A Geospatial Knowledge World

Final Report Summary - GEOCROWD (A Geospatial Knowledge World)

The goal of the GEOCROWD project (https://web.archive.org/web/20180910202612/https://www.geocrowd.eu/) was to promote the GeoWeb 2.0 vision and to advance the state of the art in collecting, storing, analysing, processing, reconciling, and making large amounts of semantically rich user-generated geospatial information available on the Web. The research work centred around three main themes: integrating geospatial content streams, geoweb data management, and accessing geospatial content. Below we provide an overview of the work that has been carried out in the project, outlining the main achievements.

The first theme deals with integrating geospatial content streams. The partners have worked in this area trying to tackle several aspects involved. In particular, NTUA worked on the development of a multi-dimensional semantic space for modelling user-generated geospatial content. Their research focused mainly on the definition, extraction, formalisation, and integration of geospatial semantics from heterogeneous information sources. The main data source for this study was travel blogs generated by non-expert users. Uni-HB focused on hybrid qualitative and quantitative spatial reasoning and analysis. Their work enhances GISs with the ability to handle queries that integrate qualitative spatial relations, such as topology, direction and distance, with qualitative spatial relations. Their approach has been applied and tested in the Bremen Tourists Advisor application and in a qualitative emergency management system. AU studied the efficient extraction of content from enriched geospatial and networked data, such as the data that is widely observed in web content. They proposed techniques for processing continuously moving top-k spatial keyword queries over spatial text data, including streaming content from social media. Moreover, they investigated the integration of spatial distance and textual relevance into a single ranking function. Furthermore, they worked on the problem of retrieving a small set of objects that best satisfy allying and alienating preferences on non-spatial properties of spatial web objects. Finally, in order to enable the interpretation and understanding of large-scale, highly networked web data, work has been conducted on the compression and summarisation of large-scale networks. NUIM and USTAN worked on the analysis of spatiotemporal patterns discovered from volunteered geographic information (VGI). They studied urban dynamics based on VGI and developed a quantitative data-driven methodology to extract semantically rich modalities and spatiotemporal patterns from streaming VGI data using probabilistic topic models. Furthermore, they worked on the use of VGI for the study of human movement patterns. Finally, they studied how spatial-temporal patterns can be extracted from VGI with spatial interaction models. In addition, USTAN studied the use of user-generated content, revealing a lack of sufficient techniques for discovering spatial knowledge and relations from it. To infer various trip-making behaviours, they are conducting an extensive experiment involving the collection of GPS traces from people in the area of Edinburgh city. Their work also focused on geospatial data fusion. Spatial statistical techniques were proposed that relate heterogeneous quantitative and qualitative data to provide an integrated view of the actual situation in the urban system. Spatial interaction models were applied in order to extract higher knowledge from VGI and user-generated content.

The second research theme deals with GeoWeb Data Management. In this area, ETH Zurich studied the interaction between mobile phones and cloud computing, and in particular how to best deal with increasing loads in order to increase responsiveness and minimise network traffic. Research showed that alternative data processing approaches are capable of dealing with very high loads, in terms of service requests, while providing performance guarantees and supporting complex queries. Research focused on efficient operators for processing data streams in a scalable and timely manner. NTUA studied the crowdsourcing of geospatial data. Geospatial data were semantically enriched with crowd-sourced data in order to capture user sentiment in the form of links-of-interest among pairs of geospatial objects that represent user choices, which serve as the basic building block for more accurate and relevant location-based-services. To this end, methods that allow the efficient querying of relevant nearest neighbour queries were investigated and proposed. Research focused on data extracted from travel blogs. FU Berlin has developed methods for crowdsourced routing. The main idea is that navigation systems can be improved by using people experiences as feedback. To this end, a routing algorithm, called CrowdRoute, that utilises crowdsourced data from mobile devices and the web has been developed. Additionally, they proposed a generic data model for crowdsourcing that is applicable to real-world crowdsourcing tasks. Moreover, they have investigate the exploitation of webtables for itinerary planning.

Finally, the third research theme focuses on accessing geospatial content. In this direction, NTUA investigated the problems of collection and quantification of user-contributed geospatial content. They studied the identification of quantitative geospatial data from user-contributed qualitative spatial relationships. In addition, they developed a method for automatic road network generation and map-attribute generation based on vehicle tracking data. They studied the problem of the extraction of spatial relations from user-generated texts, the quantification of these relations based on probabilistic methods, and the reasoning about object locations based only on the qualitative data. They integrated extracted qualitative knowledge to road networks and provided a route planning approach that accounts for the semantic closeness between pairs of points of interest. AU focused on indoor location-based services. They developed a data management infrastructure that captures indoor distance. They worked on the integration of different technologies for positioning. The concept of an organic system was investigated. An organic system is initialised and maintained by users. A prototype that integrates Wi-Fi and video cameras for positioning has been proposed. Additional research investigated the extraction of information about people's movements, the extraction of typical indoor movement patterns, and the exploration of the temporal aspect of the frequency of an indoor pattern. Uni-FB studied sensor-based data collection. They investigated the use of mobile sensors for user input, image processing technologies and intelligent annotation of geospatial features. They focused on micro-mapping, the geometric correct acquisition of small spatial entities and developed MAPIT, which based on everyday smartphone technology, enables the rapid, barrier-free acquisition and contribution of full geometric information about small spatial entities. MAPIT includes a mobile application available on Android smartphones and a web service portal deployed within a central server. Their approach has been evaluated through user usability studies. FU Berlin studied targeted alerting in emergency response situations. They developed a model that enables the representation of the state of the group of mobile users, and allows the analysis of clusters of mobile users. They evaluated different datasets for multi-user context modelling. Also, their research focused on clustering approaches for the identification of target groups in the population based on attributes, such as user location, user transportation mode, and hazard location. Finally, they outlined the role of crowdsourcing in emergency response and proposed a suitable model. ETH Zurich, in order to enable constant response time and low latency guaranties, given the the diversity of data and workloads in the multicore era, considered a hybrid architecture comprising a row store engine, a column store engine and a key-value store. In addition, they investigated the problem of efficient information delivery and personalisation of information services, particularly on data going from a cloud provider to a mobile phone, and focused on the efficient utilisation of modern hardware platforms to provide scalability to information services.

The research activities conducted within each node of the network in the themes outlined above were additionally complemented, supported and enhanced by a series of training and networking events, such as the organisation of workshops and summer schools. Most notable is the GEOCROWD international workshop (Int'l Workshop on Crowdsourced and Volunteered Geographic Information) co-located with the major event in Spatial Data Management and GIS, the ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, which has been running since 2013 under the auspices of the GEOCROWD project. Through these actions, the GEOCROWD project established a fertile environment for creating and coordinating research, education, and innovation activities, laying the foundations for permanent programs that guarantee sustainability of the scientific effort in the long term, bringing together research leaders with interrelated but complementary expertise, and raising the value of significant European innovations that exist in the area of Intelligent Information Management.