Our group’s work focuses on developing methods to aid the use of geographic information not only by specialists, but also by society in a more general sense. Doing so requires us to, for example, develop basic spatial algorithms, apply computational text analysis or do ethnographic fieldwork. Thus, for instance, we investigate how people describe the world around them and seek to develop methods that link computational representations to these descriptions.
Extracting geosemantic information
We are interested in how people describe the world and how such local conceptualizations of space can be documented and used to improve current representations in geographic information systems. One approach to this problem is to extract information about vernacular place names and landscape descriptions from a range of different sources such as unstructured text, social media and ethnographic data.
- How do people from different cultures differ in the way they conceptualize and refer to places and landscape features?
- How can we automatically extract information on how people describe their environment (e.g. place names and geographic features) from textual descriptions?
Does geography matter when we search for information? We seek to develop methods to retrieve and rank documents (e.g. texts, images) from web pages, news archives, or image collections. By linking the geographic information that is implicitly included in these data in ‘Geographic Search Engines’ we aim to improve geographic search so that users will find the requested information more easily.
- How can geographic information improve the answers from search engines on a users’ query?
- How can textual descriptions, for instance in web pages, be linked to space?
What influence do different representations of the environment in geographic information have on the results of physical models? We evaluate methods to represent, interpret and make use of terrain data. Furthermore, we seek to understand uncertainties in products that are derived from digital terrain data, for instance hydrological models.
- How do resolution and uncertainty in terrain data influence modelling results?
- How can landform types, such as valley or mountains, be extracted from terrain data?