Exploratory Search in Databases and the Web

The traditional way a user interacts with a database system is through queries. Structured query languages, such as SQL for relational data, XQuery for XML, and SPARQL for RDF data, allow users to submit queries that may precisely capture their information needs, but users need to be familiar with the underlying ontology and data structure and of course the query language itself. Moreover, users need to some extent be familiar with the content of the database and have a clear understanding of their information needs. These requirements stand as the weaknesses of this interaction mode. As data stored in databases grows in unprecedented rates and becomes accessible to diverse and less technically oriented audience, new forms of data exploration and interaction become increasingly more attractive. The World Wide Web represents the largest and arguably the most complex repository of content. Users seek information on the web through two predominant modes: by browsing or by searching. In the first mode, the interaction between the user and the data repository is driven directly by the user’s interpretation of their information need and their information foraging constraints. In the latter mode, a search engine typically mediates the user-data interactions and the process starts with the user entering query-terms that act as surrogates for the user information goals. Freetext queries allow end-users a simple way to express their information needs independently from the underlying data model and structure, as well as from a specific query language. Given a query, the most common strategy has been to present the results as a ranked list. Users have to subsequently peruse the list to satisfy their information needs through browsing the links and/or by issuing further queries. However, the information in the web gets rapidly diversified both in terms of its complexity as well as in terms of the media through which the information is encoded, spanning from large amounts of unstructured and semi-structured data to semantically rich available knowledge. Increasing de-