Future Work

Our goal is to make another order of magnitude increase in the size of the datasets that DQIs can deal with (1,000,000 records with 10 attributes). We plan to:

• implement other widget types, e.g., alphanumeric sliders.
• try spatial data structures like k-D trees to see how they effect the times for selection and querying. (As a general non-worst-case rule of thumb, spatial data structures answer range queries in time where H is the set of hits and a is the number of attributes in the input dataset. This could be good for selection, because it is sub-linear. But it could be bad for querying, because it is close to linear, and prior work seems to confirm this doubt [10, 11, 13]. Instead, we will use an incremental approach where we compute the difference between consecutive hit sets, which in practice will take time only .)
• combine our DQI with a query previewer [3] in order to produce a new state of the art in interactive dynamic database access.