Turning DTDs into specialized tree-automata-based schemata to match a collection of marked-up documents

Regular tree automata (RTA) or, equivalently, forest regular grammars (FRG) have been re ently proposed for use as XML (extended markup language) s hemata. They are more powerful than usual XML DTDs (do ument-type definitions), make the implementation, optimization and pruning of XML queries easier and allow for the implementation of ontext-sensitive ontent models. We des ribe a method for the automati generation of a spe ialized RTA-based s hema from a sour e DTD and a sample of marked-up do uments showing ontext-sensitive behaviour in ontent models. It reates the smallest RTA-based s hema with whi h all the XML do uments in the sample omply and whi h does not a ept any do uments not valid a ording to the original DTD. In this way, new les an be reated, parsed, and queried using the spe ialized s hema but still be ompliant with the original DTD. The tool is urrently being tested at the Miguel de Cervantes digital library at the University of Ala ant (http:// ervantesvirtual. om).