Science.gov goes 3.0
- By Joab Jackson
- Nov 21, 2005
The newest version of the Science.gov
search engine uses metadata to rank results, which should give users more accurate responses to queries, according to the Science.gov Alliance.
'We're trying to plow new ground with searches among multiple databases,' said Walter Warnick, director of the Department of Energy's Office of Scientific and Technical Information, which hosts the Science.gov site. Warnick said the new version of the search engine makes searches for obscure terms easier, thanks to the new technology.
Version 3 of Science.gov, which went operational last week, incorporates bibliographic information into its searches. The feature, dubbed 'MetaRank,' ranks the relevance of data to the search term by using title, author, date, abstract and other keyword identifiers. It places those data sources that appear to be closest to what the user wants highest in the list of results.
Although MetaRank performs relevancy ranking much like the Google commercial search engine does, the two approaches are entirely different, according to Warnick. Google uses Web crawlers and a ranking technique called PageRank that analyzes the links among Web pages. MetaRank was developed primarily for database data, not Web pages. As such, it uses other algorithms, such as those that gauge how often the search term appears in the abstract and where it is in the abstract's text.
'MetaRank greatly expands the amount of information that comes back with each hit, so the relevance ranking can be more sophisticated,' Warnick said.
Version 3 also enjoys a number of other new features. Boolean searching has been modified so that the rules more closely resemble those used by commercial search engines. Users can now search in specific fields, such as dates or author. Another new feature displays immediate views of data before the results of the search are fully returned.
Introduced in 2002, Science.gov offers the public a unified search service for governmental scientific information. It searches across 30 databases and 1,800 Web sites.
Deep Web Technologies LLC of Los Alamos, N.M., developed MetaRank through a grant award from the Small Business Innovation Research Program, as well as from contributions from participating agencies of the Science.gov Alliance.
The alliance is composed of 17 organizations within 12 federal science agencies, including scientifically oriented organizations within the departments of Agriculture, Commerce, Defense, Education, Energy, Health and Human Services, and Interior, as well as the Environmental Protection Agency, NASA, the National Science Foundation and the Government Printing Office.
Joab Jackson is the senior technology editor for Government Computer News.