DBExplorer investigates the use of keyword search in relational databases.  It mentions several interesting problems that are encountered in keyword search as opposed to structured queries. For example, SQL applications require knowledge of schema and the keyword search doesn’t. Secondly, given a set of keywords, a match is found by joining several tables on the fly. Thirdly, Indexes need to be leveraged for efficient keyword search. Fourthly, common data structures used for document collections such as inverted lists are replaced by symbol table. For each keyword, a list of rows is kept in the symbol table that contains the keywords. Alternatively, for each keyword, a list of columns can be kept that contains the keywords.  Search is performed across multiple tables using a graph where the nodes are the tables and the edges are the foreign-key relationships. When looking for a co-occurrence, the tables are joined on the fly by exploiting the schema as well as the content.