SEM: mining spatial events from the web

  • Authors:

  • Kaifeng Xu;Rui Li;Shenghua Bao;Dingyi Han;Yong Yu

  • Affiliations:

  • Department of Computer Science & Engineering, Shanghai Jiao Tong University, Shanghai, P.R. China;Department of Computer Science & Engineering, Shanghai Jiao Tong University, Shanghai, P.R. China;Department of Computer Science & Engineering, Shanghai Jiao Tong University, Shanghai, P.R. China;Department of Computer Science & Engineering, Shanghai Jiao Tong University, Shanghai, P.R. China;Department of Computer Science & Engineering, Shanghai Jiao Tong University, Shanghai, P.R. China

  • Venue:
  • PAKDD'08 Proceedings of the 12th Pacific-Asia conference on Advances in knowledge discovery and data mining
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper is concerned with the problem of mining spatial events from the general Web. General search engine is inconvenient when searching vertical information (e.g., locations, experts) since it is designed for general purpose. For example, when finding the battlefields of World War II, listing the Web pages by relevance is not enough to tell users the spatial information clearly. A categorized result along with a map indicating these battlefields would be much easier to read. To present such a result, we propose a novel algorithm called Spatial Event Miner (SEM) to mine spatial event information from the general Web. Given a simple keyword query, SEM first collects and ranks a set of relevant locations from the Web. Then, to describe the events happened in the collected locations, SEM detects and sums up salient phrases as event topics from the context of these locations. For each specific location, the hottest event topics are also listed for quick understanding. Finally, a clear spatial distribution on the events of a given query is presented to the users. A prototype system based on SEM is also implemented. Preliminary experimental results on a set of 40 queries show that the proposed approach can capture the spatial event information effectively.