Abstract
Though deep web search research has a long history, no convenient search engine has materialized of late to support access to the deep web databases in ways similar to shallow web engines such as Google or Bing. Surfacing and custom access engines are the two most common techniques available to search a tiny fraction of the deep web today with the concomitant limitations. In this paper, we discuss the implementation of DeepIR as a possible deep web search engine. It is powered by a flexible declarative query language called DQL that is similar to SQL. We assemble and leverage existing technologies to realize DeepIR and identify possible pain points for further research. We believe investigating the identified technological hurdles will help develop a smarter deep web search engine. The advantage of DeepIR is that its system components can be replaced with more powerful ones to improve its performance without impacting its query semantics.