Geoffrey Hayes
   Professor Jianbo Shi

There exists an emerging gap in the use of the Internet to approach general location-based event searches for web users. While many search engines and geographical service websites allow users to lookup points of interest by location (e.g. Citysearch, Google Maps), these fail to address the issues of locality, punctuality, and user-preference. WhatsHere is a web application allowing a community of users to search for and contribute to localized events.

The system uses a crawler to mine data from newspapers and websites across the Internet. Based on a collaborative recommender system utilizing context information, the WhatsHere system takes a novel approach by focusing strongly on source similarity and adaptive application-specific context curves for individual users. Our algorithms provide linear-time queries to ensure scalable real-time results.

We have developed a search algorithm where a web user may search events by simply specifying a lookup location. Combining this address with the user's history and preferences, the system delivers a sample of results to the client. The algorithm then uses the interactions of the web user in order to drill-down on the intent of the user for future searches. Having observed trends that a single source (web site, newspaper, ...) often hosts by-and-large similar events, WhatsHere categorizes events by their originating source.