Evaluating a Faceted Search Index for Graph Data

Abstract

We discuss the problem of implementing real-time faceted search interfaces over graph data, specifically the “value suggestion problem” of presenting the user with options that makes sense in the context of a partially constructed query. For queries that include many object properties, this task is computationally expensive. We show that good approximations to the value suggestion problem can be achieved by only looking at parts of queries, and we present an index structure that supports this approximation and is designed to scale gracefully to both very large datasets and complex queries. In a series of experiments, we show that the loss of accuracy is often minor, and additional accuracy can in many cases be achieved with a modest increase of index size.