Geographic Metadata Searching with Semantic and Spatial Filtering Methods

Web search engines, such as from Google, are very good at finding relevant information in documents and web pages. However, when such tools are used to find spatial web services, the user has to be very specific in describing what they are looking for to find relevant results in high-ranked positions. To locate an Open Geospatial Consortium-compatible web service relating to soil in Australia, a query such as “getcapabilities australia soil” is required to find relevant results, as there are no spatial constraints available. Current spatial data discovery systems, such as spatial catalogue systems, generally keyword match user queries to the content of metadata catalogues. Such systems also provide basic spatial constraints, which limit the user’s ability to find results. A combination of semantic and spatial search techniques are required to effectively search geospatial data, as existing systems are primarily designed to search human-readable documents. A search algorithm is presented which uses such techniques to expand text queries to find more relevant spatial datasets through spatial filtering, natural language query decomposition and the use of thesaurus graphs to expand queries. A Resource Description Framework (RDF) schema that extends the ISO 19115 specification is explored as part of the query expansion technique, including the evaluation of tools to generate these graphs from unstructured documents which allows the overcoming of restrictions to access data behind an organisation’s firewall. A prototype has been written as a web application, using the Django framework and the Python programming language and the Natural Language Toolkit (NLTK) interface to WordNet. Initial tests on seperate components of the system as well as the above system has shown the feasibility of the search system as a whole.