Answers are grounded in this paper and cite their sources.
We consider the problem to infer a concise Document Type (DTD) for a given set of XML-documents, a problem basically reduces to learning of concise regular from positive example strings. We identify such classes: single occurrence regular expressions (SOREs) and chain regular expressions (CHAREs). Both capture the far majority of the regular expressions occurring practical DTDs and are succinct by definition. We the algorithm iDTD (infer DTD) that learns SOREs strings by first inferring an automaton by known techniques then translating that automaton to a corresponding , possibly by repairing the automaton when no SORE can be found. In the process, we introduce novel automaton to regular expression rewrite technique is of independent interest. We show that iDTD existing systems in accuracy, conciseness and . In a scenario where only a very small amount of data is available, for instance when generated by Web requests or by answers to queries, iDTD produces expressions which are too specific. Therefore, we introduce novel learning algorithm crx that directly infers (which form a subclass of SOREs) without going an automaton representation. We show that crx very well within its target class on very small data . Finally, we discuss incremental computation, noise, predicates, and the generation of XML Schemas.
No takes yet. Share an insight, caveat, or question.
Bex et al. (2006) studied this question.