Semantics, Reasoning and Cognition
Theme Overview:
This work involves the development of methods to represent the behaviour of geospatial objects and geospatial functionality. Such methods may be used to assist in the interpretation and application of geospatial information, including the automation of matching geospatial functionality (e.g. web services) with user objectives, and automatic geospatial analysis. The research is particularly focussed on the application of these methods in the context of time-critical situations (for example, emergency management) and issues of social and humanitarian importance.
Recent work focuses on the use of linguistic semantic primitives from Natural Semantic Metalanguage (NSM) to express the semantics of spatial objects, relations and their behaviour and to convert these NSM representations into formal representations in information systems. Forthcoming publications include an examination of the way people describe spatial relations using NSM (with human subject experiments) and the definition of transformations and relationships between different spatial primitives from NSM.
Additional research involves the development and use of knowledge infrastructures to support the use of geographic information by scientists in the scientific process, and involves various collaborations with Edinburgh, Cardiff and Muenster Universities, with particular focus on the evolution of knowledge and discovery and visualisation/representation of patterns in relationships between knowledge artefacts.
Parallel to this we are undertaking research in vernacular geography. Investigation into our cognition of space and methods for capturing these ideas in computer systems is required to allow vaguely perceived areas to be analysed with computerised tools. Vernacular geography refers to colloquial place names and vague geographical extents used by people on an everyday basis. These frequently provide alternative names to the official ones, and may only map poorly onto the officially recognised boundaries. Individuals often refer to places with no formal definition, as for example in the sentence “I’m going into town”. If an official designation exists for an area, such as an administrative boundary, it is unclear as to if or how much an individual’s perception differs; studies suggest that while administrative boundaries influence part of people’s perceptions this is alongside other spatial factors.
The issue of improving geographic information system (GIS) efficiency is becoming particularly apparent with the growth of the Internet. It has been estimated that between 13-15% of search engine queries contain a place name and this has prompted a need for suitable mechanisms capable of retrieving information from the unstructured Web. Such mechanisms are considered part of the domain of geographic information retrieval and require the ability to process vague geographic terminology. This ability of systems to consider vagueness also extends to the development of interfaces based on a more natural language.
Much of the ongoing research in retrieving geographic information from the Web uses text documents. Most of these approaches view the Web as a large pool of data collected for other purposes that can be harvested or mined for place-related content by a third party.
However, we believe that an alternative is possible where participants directly contribute gazetteer data but as a secondary purpose to the Web system. We also recognise that vernacular geographies do not have precise boundaries – the boundaries may be vague for any one person and the opinions on boundaries will vary from one person to another. Our aim therefore is to allow the regions collected to be modelled as having fuzzy areal definitions.
Research questions:
-
Can vague areas be actively elicited from a social media application to create a gazetteer?
-
How do such areas compare with those obtained from mining other geotagged data and official definitions such as those captured by the OS? How does the social media source type affect the boundaries derived?
-
How do we use the vague areas in computer-aided decision making? Do we need to resolve the areas to Boolean boundaries or can we use fuzzy logic?
-
What more can the degrees of fuzziness tell us about the cognitive processes that define place?
A third topic of work explores the interaction between spatial cognition and language. In this topic we are exploring the language of navigation with colleagues from the School of English Studies. We are using methods from applied linguistics to explore the way in which people ask for and give directions. This explores different transport modes (as a pedestrian, or by bus or car for example). We are not only interested in the geographical objects referenced in the directions but also the form of directions (“third on the left”, “100m on the left”, “Side Road”, “left by the church”), and more broadly the form of language used in the communication. By understanding the language of navigation we can both better inform the geographical databases used to support automated direction-giving and the manner in which directions are described to a user.
CGS projects involving Semantics, Reasoning and Cognition:
Events:
CGS staff focussing on this theme:
-
Jeremy Morley
-
Kristin Stock
-
Claudia Cialone