In order to retrieve an item of information on the Web, many search engines have been proposed. They are rarely efficient at the first attempt: the display of results “forces” the user to navigate. In parallel, Web query languages have been developed to avoid these two sequential phases: research then navigation. In this context, the QIRI@D experimental platform, based on the functional programming language SgmlQL, enables both information retrieval and manipulation of distributed semi-structured documents published on a sub-network made up of the sites where QIRI@D is running. It is possible in an unique query to specify criteria to find a document, to filter it, to extract parts of it for building the result. An automatical enrichment of any published document is used to improve the search efficiency.