RSS Feeds

Email a Friend Printer Friendly

Adding more meaning from place searches

February 14, 2005

Trawling the web for place-related information is tedious at the best of times. A new search engine, being tested in Europe, recognises geographical terminology and has the intelligence to understand the searches and match them to places.

Punch in a request for all restaurants 'near' a particular town, and the average search engine will draw a blank. "The problem," says Christopher Jones, coordinator of the IST project Spirit, "is they do not understand place-specific requests. They find documents with the place names, but don't know what they are or what is nearby."

Web engines offering to search for information by geographical location, he adds, are experimental or link responses only to business directory listings. Which is why five European universities and a national mapping agency joined forces to produce a 'spatially aware information retrieval' system, for use with the Internet.

Within 18 months, the project had a prototype. Based on a pure text search engine developed by project partner Sheffield University, 'Spirit' specialises in finding information that relates to a particular geographical location.

"Our demonstrator focuses on Europe, specifically cities such as Cardiff, Edinburgh and Zurich," says Jones. "It contains a test collection of around 20,000 documents, expanding soon to around 100 million, built on a terabyte of web data."

From April 2005, anyone can test the demonstrator by logging onto the project website. Adds the coordinator: "We offer access to restricted geographical areas, partly due to copyright issues. But testers will discover that Spirit is fast and effective for certain types of query."

Because Spirit is specialised for information that is geographically specific, users can find web pages relating to particular places with ease. It understands what people mean when they search for data that is spatially related to a named place, such as being 'north of' or 'near' or 'inside' the place. For example, it can find places within places, examples being the boroughs of London, without the user having to specify their names in the original search.

At the heart of Spirit is the geographic ontology, a knowledge base of geographic places built from various resources. "This base is vital for geotagging or characterising documents according to their geographical context. It is also essential for recognising the presence of place names in user queries," says Jones. The search engine can, for instance, alert the user if a place name has multiple occurrences and prompt the user to say which one they are interested in - very useful for tracking the one in ten places that share a name with somewhere else. Relevance ranking also has geographical intelligence and sorts the results of queries according to geographic relevance, as well as with respect to the textual subject matter.

The project's demonstrator is being refined with improved spatial indexing and geotagging. The partners will also extend the geographic ontology to recognise imprecise and vague names, such as England's Lake District or the South of France.

"We seek a promoter to take this software beyond the project's lifetime," says Jones. He believes people working in his own discipline of Geographical Information Systems would benefit from Spirit, since GIS users often do specific searches based on names rather than map coordinates.

Other commercialisation targets include digital libraries and search-engine companies expanding their conventional search methods. Lastly, according to the coordinator, Spirit could open up new markets for location-aware mobile devices, allowing users to find geographically specific web resources associated with their current location.

IST Results