IBM Research - Almaden - User System Ergonomics Research

 FIRE - Query Refinement


  Finally, when the searcher wishes to issue a new refined search, the system includes its term suggestions automatically but allows the searcher to delete or modify them according to her understanding of the query. This step of the dialog takes place in the same form as the initial graphical query.

In this example, the FIRE engine has included an addition set of terms including ibms, microsoft, operating, and industry. The searcher has determined that industry is not useful. Therefore she deletes that term before re-submitting the query.

This modified query is sent to the search engine and the process is continued. Note that the documents which were previously marked as "relevant" will automatically appear in the relevant list so that the searcher need not look through them again. Likewise, any trashed documents which are still returned will be hidden in the trash can so the searcher will not be bothered with them.

We have found that the FIRE query refinement system provides significant aid for textual information retrieval. Using its suggestions usually improves the precision of a search by 30% without significantly reducing recall. But more importantly, the system helps searchers understand their queries better and stimulates their thinking by providing hints as to the differences between relevant and trash documents.

        width= -----
  | USER - Home Page | USER - Projects | Computer Science - Home Page |
    Ron Barber
    IBM Almaden Research Center
    650 Harry Road
    San Jose, CA 95120-6099

Last updated: Thursday, 21-Jul-2011 15:25:04 PDT

[ Almaden Research Center | IBM Research ]
IBM home page | Order | Search | Contact IBM | Help | (C) | (TM) ]