Bergman (2001) is an extensive study of the deep Web.
Even though this study is old by web standards, it shows how sampling through search engines can be used to help estimate the amount of unindexed content on the Web.
This study estimated that 550 billion web pages existed in the deep Web,
compared to 1 billion in the accessible Web. He et al. (2007) describe a more recent survey that shows that the deep Web has continued to expand rapidly in recent years.
An example of a technique for generating searchable representations of deep Web databases, called query probing, is described by Ipeirotis and Gravano (2004).