Spatial and Thematic Ontology in Cultural Heritage Information Systems

This project investigated the design and implementation of a knowledge-based system for thematic and spatial access to information about archaeological artefacts. The project explored the use of an integrated spatial and thematic ontology to address many problems related to the search and retrieval of archaeological information. One common problem is the frequent mismatch between the terminology employed by users to access the information, and that used by developers to index it. Systems may lack the capability to process queries with loose information demand, since they are unable to imprecisely match query terms. Another challenging area is that of handling the spatial information associated with archaeological artefacts. Places normally have different version of names, often change in size, boundaries, and centroid co-ordinates. Generating appropriate spatial footprints to adequately represent their extents and infer their spatial relationships forms another challenging area. Several methods and approaches were investigated in this project to overcome the above problems. The ontology was designed to handle different versions of place names and ease the terminology problem by controlling vocabulary using integrated spatial and thematic thesauri. Semantic distance measures were employed to imprecisely match query terms, providing ranked lists of similar objects. A thematic measure was developed based on semantic-path traversals to expand artefact types. The project explored some of the main issues affecting the use of associative thesaurus relationships in query expansion. One approach found useful was to specialise these relationships and select which ones to activate based on the query context. The spatial domain in the ontology was constructed from a geographical thesaurus, enriched with spatial relationships. Spatial distance measures were used to provide more flexible retrieval for queries with spatial content. The two main spatial measures developed in this project were based on some of the common spatial information provided by most gazetteers, such as centroid co-ordinates and administrative hierarchies. Place co-ordinates were used to measure similarity of places according to Euclidean distances, while the hierarchical associations of places were useful to measure place similarity when administrative divisions are important or when co-ordinates are not available. A dynamic spatial approximation method was developed that uses sparse spatial information to generate approximated boundaries, and can be employed to infer topological, proximity, and directional information.