What Are We Searching For? Analyzing User Objectives When Searching Relational Data
Joel Coffman, Alfred C. Weaver · 2012
Many sources of information available online cannot be accessed using Internet search engines, yet their users desire the same high-quality search results that they have come to expect when searching the web. Keyword search techniques for relational databases hold promise for lling this void, but existing ad-hoc evaluations pose a major barrier to improvement in this eld. In this paper, we investigate the types of queries that users might submit to a relational keyword search system. We develop a framework for analyzing queries in this context, which is similar to existing taxonomies for web searches. Our analysis of search engine query logs reveals considerable variation among queries for dierent datasets. We show how to use our framework to create representative query workloads to evaluate relational keyword search techniques and illustrate the dierences between our workloads and those used previously in the literature. This work closes an important gap in the existing evaluation methodology and promotes continued improvement to relational keyword search techniques.