Review

This paper offers a new user interface for document retrieval systems in which an array of 100 cells is marked with the strength of agreement with a particular query phrase, Boolean expression, or cluster. Users can flip quickly between different queries or clusters; the position of each document in the cell array remains the same, helping the user note which documents seem to be strongest. I was surprised that they didn't use some kind of overlapping or see-through technique; for 100 cells in a window of at least 100K pixels, there is an opportunity to provide more than just a color and intensity level for the cell. It is not clear how the cells are ordered; this is another opportunity to do something to improve presentation effectiveness.

They discuss a brief evaluation which showed that this technique works better than simple ranked retrieval, but with little detail on how the relevance assessments or queries were chosen. They also explain how to avoid loading down the search engines (or requiring changes to them) by using a special presentation engine in front of a variety of search engines.

We need ways of using graphics more effectively to improve information retrieval, but I find it hard to evaluate how big a step forward the authors' grids would be. Although they seem to improve over simple lists, or documents, I am not convinced they improve over systems like Envision or Tilebars.