Sorting out searching
Ben Shneiderman, Donald Byrd, W. Bruce Croft · Communications of the ACM · 1998
The ideal user interface is comprehensible, predictable, and controllable, but many current textsearch interfaces-especially on the World-Wide Web-involve unnecessarily complex and obscure features.The result is confusion and frustration for advanced users as well as for beginners, scientists, and students [8].Even when a user interface's design is improved, inconsistencies can cause mistaken assumptions and increase the likelihood of failure to find relevant documents as users move from one search system to another.For example, the search string "Hall effect" could produce various searches, including:• Exact match for "Hall effect" • Case-insensitive match for "hall effect" • Best match for "Hall" and "effect" • Boolean match for "Hall" and "effect" • Boolean match for "Hall" or "effect" Few systems spell out the interpretation they are using.Furthermore, systems often use surprising query transformations, unpredictable stemming algorithms, and mysterious weightings for fields.And in many systems, the results are displayed in a relevance ranking whose meaning is a mystery to many users (and sometimes a proprietary secret).Our four-phase framework increases clarity and S u p p o r t i n g T