The "ranked list" problem is well known by search engine users: the long list of results retrieved by a web query discourages users, that usually look at the first one or two pages only, while useful inforation could be hidden in results that are not in the top posiotions.
In this talk, we will present the meta-search system named Matrioshka.
This system is based on clustering of retrieved documents, and provides several features that allow users to conduct complex exploratory activities.
Besides the others, recent developments added a set of "Soft" operators to the system, so that groups of documents can be combined based on approximate matching of the content of retrieved documents.