Introduction

Dynamic queries have been shown to maintain advantages over traditional methods of information retrieval. [2,3,4] Applications of dynamic queries include searching large datasets in networked environments(NASA EOSDIS) [6,7] and a real-estate "home-finder" which may allow agents or clients to find real-estate that matches certain criteria as chosen through the dynamic query interface. [4]

Our intent was not to discover whether dynamic queries work better than other information retrieval methods. We assumed these advantages (properties) of dynamic queries: [2]

  • visual representation of the world of action including both the objects and actions
  • rapid, incremental and reversible actions
  • selection by pointing (not typing)
  • immediate and continuous display of results

    We wanted to find out a little of what determines success when dealing with a dynamic query interface. An inherent problem with filtering data from a large data set is in choosing the query criteria. [5]    The interface we chose to work with in looking at this problem is called Spotfire.

    Spotfire is a "visual data mining" program which allows users to browse, search, and analyze data in a direct manipulation interface. Of the tasks a user is able to accomplish with Spotfire, he or she may want to quickly locate or identify a single item(or more) from a data set. Using the program's dynamic query functions, a user filters out unwanted or irrelevant data, leaving only the data which is "near" the desired data. Figure 1 below shows the Spotfire interface.

    We have a simple data set, for example, a database of movies and films with characteristics such as year of release, length, subject, popularity, title, etc. Given a 2-D graph of the length(y-axis) of films by year(x-axis), we can easily find the longest film by glancing at the static bar graph. However, to quickly reach a set of desired data in a data set with many characteristics, as in Spotfire, how do we decide which criteria to filter and in what order? In our example, what is the most efficient way to find the most popular film(s) in the drama genre which is also between 1 1/2 to 2 hours long? We wanted to measure user performance with different manipulations (i.e. setups of characteristics on the starfield display) of the 2-D starfield display and query devices.

    The practical real-world applications of dynamic query interfaces like Spotfire are many. Spotfire has already manifested itself in areas such as pharmaceutical research and other industry sectors.[10]   One area we hope to see application in is the retail sector. A film database such as the one we experimented with could be evaluated for video rental businesses. The other area of application would be for organizations, libraries, or companies dealing with historical archival and statistics.

    Figure 1
    (back to text)
    Continue --> Experiment

    Return to Main
    Return to SHORE '98