Geographic search engines

As the World Wide Web grows rapidly, web search engines are becoming indispensable tools for internet users to satisfy their information needs. In many cases, people are interested in information with geographic constraints, such as tourism information, local businesses, or locally relevant events. Geographic web search engines allow users to constrain and order search results by focusing a query on a specified geographic region. In this dissertation, we describe the architecture and major components of a geographic web search engine and address the challenges of integrating geographic information into a keyword-based search engine. A web search engine performs three basic functions: crawling, index construction, and query processing. During crawling, web pages are downloaded and parsed for hyperlinks. After a sufficient number of pages has been acquired, a text index structure is built on the collection. When users issue queries, the top results are retrieved by traversing the index structure and ranking encountered web pages. Building a geographic web search engine involves additional challenges, including geographic feature extraction and modified query execution strategies. There are many different sources of information about the geographic properties of web documents that can be exploited for data extraction. For example, geographic references can be extracted from documents and URLs, or from the whois entries of the sites. For each page, the collection of all regions that the document provides information about is then stored in a structure called a geographic footprint. Once these initial geographic footprints have been created, they can be enhanced by propagating their information across hyperlinks and site structure. Geographic search engine query processing is different from standard search engines in that it requires a combination of text and spatial data processing techniques. Given a query, it identifies pages that contain the keywords and whose page footprint intersects with the query footprint, and ranks these results according to a combination of a term-based and a geographic ranking function. We propose several algorithms for efficient query processing in geographic search engines, integrate them with a text query processor, and evaluate performance of the algorithms on large sets of real data and query traces.