Restaurant search with predictive multispace queries

This paper describes a web-based search application used for locating restaurants in a multidimensional content space via interactive space visualization. The primary goal of our research is to reduce the cognitive complexity of query selection, as well as to help the user avoid one of the common pitfalls of traditional search mechanisms: retrieving too many or too few results. Our method approaches this problem by continuously staying one step ahead of the user, constructing a graphical output to summarize how further query adjustments will impact subsequent search results. In actively precomputing queries ahead of the user we help them avoid the trial-and-error search process often associated with trying to find something in an unfamiliar, opaque database. We combine our visual search method with a profile-based knowledge system which helps users find restaurants accessed by people with similar interests.