Retrieving News Stories from a News Integration Archive

摘要

The distinctive features of the Bveritas online news integration archive are as follows: 1) automatic clustering of related news documents into themes; 2) organization of these news clusters in a theme map; 3) extraction of meaningful labels for each news cluster; and 4) generation of links to related news articles. Several ways of retrieving news stories from this Bveritas archive are described. The retrieval methods range from the usual query box and links to related stories, to an interactive world map that allows news retrieval by country, to an interactive theme map. Query and browsing are mediated by a Scatter/Gather interface that allows the user to select interesting clusters, out of which the subset of documents are gathered and re-clustered for the user to visually inspect. The user is then asked to select new interesting clusters. This alternating selection/clustering process continues until the user decides to view the individual news story titles.