Curated databases are databases that are populated and updated with a great deal of human effort. Most reference works that one traditionally found on the reference shelves of libraries – dictionaries, encyclopedias, gazetteers etc. – are now curated databases. Since it is now easy to publish databases on the web, there has been an explosion in the number of new curated databases used in scientific research. The value of curated databases lies in the organization and the quality of the data they contain. Like the paper reference works they have replaced, they usually represent the efforts of a dedicated group of people to produce a definitive description of some subject area. Curated databases present a number of challenges for database research. The topics of annotation, provenance, and citation are central, because curated databases are heavily cross-referenced with, and include data from, other databases, and much of the work of a curator is annotating existing data. Evolution of structure is important because these databases often evolve from semistructured representations, and because they have to accommodate new scientific discoveries. Much of the work in these areas is in its infancy, but it is beginning to provide suggest new research for both theory and practice. We discuss some of this research and emphasize the need to find appropriate models of the processes associated with curated databases.

An ad hoc data source is any semistructured data source for which useful data analysis and transformation tools are not readily available. Such data must be queried, transformed and displayed by systems administrators, computational biologists, financial analysts and hosts of others on a regular basis. In this paper, we demonstrate that it is possible to generate a suite of useful data processing tools, including a semi-structured query engine, several format converters, a statistical analyzer and data visualization routines directly from the ad hoc data itself, without any human intervention. The key technical contribution of the work is a multi-phase algorithm that automatically infers the structure of an ad hoc data source and produces a format specification in the PADS data description language. Programmers wishing to implement custom data analysis tools can use such descriptions to generate printing and parsing libraries for the data. Alternatively, our software infrastructure will push these descriptions through the PADS compiler and automatically generate fully functional tools. We evaluate the performance of our inference algorithm, showing it scales linearly in the size of the training data — completing in seconds, as opposed to the hours or days it takes to write a description by hand. We also evaluate the correctness of the algorithm, demonstrating that generating accurate descriptions often requires less than 5 % of the available data. 1.

We study the succinctness of the complement and intersection of regular expressions. In particular, we show that when constructing a regular expression defining the complement of a given regular expression, a double exponential size increase cannot be avoided. Similarly, when constructing a regular expression defining the intersection of a fixed and an arbitrary number of regular expressions, an exponential and double exponential size increase, respectively, can in worst-case not be avoided. All mentioned lower bounds improve the existing ones by one exponential and are tight in the sense that the target expression can be constructed in the corresponding time class, i.e., exponential or double exponential time. As a by-product, we generalize a theorem by Ehrenfeucht and Zeiger stating that there is a class of DFAs which are exponentially more succinct than regular expressions, to a fixed four-letter alphabet. When the given regular expressions are one-unambiguous, as for instance required by the XML Schema specification, the complement can be computed in polynomial time whereas the bounds concerning intersection continue to hold. For the subclass of single-occurrence regular expressions, we prove a tight exponential lower bound for intersection.

Inferring an appropriate DTD or XML Schema Definition (XSD) for a given collection of XML documents essentially reduces to learning deterministic regular expressions from sets of positive example words. Unfortunately, there is no algorithm capable of learning the complete class of deterministic regular expressions from positive examples only, as we will show. The regular expressions occurring in practical DTDs and XSDs, however, are such that every alphabet symbol occurs only a small number of times. As such, in practice it suffices to learn the subclass of regular expressions in which each alphabet symbol occurs at most k times, for some small k. We refer to such expressions as k-occurrence regular expressions (k-OREs for short). Motivated by this observation, we provide a probabilistic algorithm that learns k-OREs for increasing values of k, and selects the one that best describes the sample based on a Minimum Description Length argument. The effectiveness of the method is empirically validated both on real world and synthetic data. Furthermore, the method is shown to be conservative over the simpler classes of expressions considered in previous work.

Whether beloved or despised, XML Schema is momentarily the only industrially accepted schema language for XML and is unlikely to become obsolete any time soon. Nevertheless, many nontransparent restrictions unnecessarily complicate the design of XSDs. For instance, complex content models in XML Schema are constrained by the infamous unique particle attribution (UPA) constraint. In formal language theoretic terms, this constraint restricts content models to deterministic regular expressions. As the latter constitute a semantic notion and no simple corresponding syntactical characterization is known, it is very difficult for non-expert users to understand exactly when and why content models do or do not violate UPA. In the present paper, we therefore investigate solutions to relieve users from the burden of UPA by automatically transforming nondeterministic expressions into concise deterministic ones defining the same language or constituting good approximations. The presented techniques facilitate XSD construction by reducing the design task at hand more towards the complexity of the modeling task. In addition, our algorithms can serve as a plug-in for

Although the advent of XML Schema [25] has rendered DTDs obsolete, research on practical XML optimization is mostly biased towards DTDs and tends to largely ignore XSDs (some notable exceptions non-withstanding). One

Regular Expressions (REs) form the basis of most XML type languages, such as DTDs, XML Schema types, and XDuce types (Thompson et al. 2004; Hosoya and Pierce 2003). In this context, the interleaving operator would be a natural addition to the language of REs, as witnessed by the presence of limited forms of interleaving in XSD (the all group), Relax-NG, and SGML. Unfortunately, membership checking for REs with interleaving is NP-hard in general. We present here a restricted class of REs with interleaving and counting which admits a linear membership algorithm. This restricted class is known to be expressive enough for the vast majority of the content models used in real-world DTDs and XSD schemas; moreover, we have proved in (Ghelli et al. 2007) that the same class admits a polynomial algorithm for subtyping and typeequivalence, problems which are EXPSPACE-complete for the full language of REs with interleaving. We first present an algorithm for membership of a list of words into a RE with interleaving and counting, based on the translation of the RE into a set of constraints. We generalize the approach in order to check membership of XML trees into a class of EDTDs with interleaving and counting, which models the crucial aspects of DTDs and XSD schemas. Finally, we extend the approach to REs with intersection. 1.

XML Schema Definitions (XSDs) can be adequately abstracted by the single-type regular tree languages. It is wellknown, that these form a strict subclass of the robust class of regular unranked tree languages. Sadly, in this respect, XSDs are not closed under the basic operations of union and set difference, complicating important tasks in schema integration and evolution. The purpose of this paper is to investigate how the union and difference of two XSDs can be approximated within the framework of single-type regular tree languages. We consider both optimal lower and upper approximations. We also address the more general question of how to approximate an arbitrary regular tree language by an XSD and consider the complexity of associated decision problems.

To experimentally validate learning and approximation algorithms for XML Schema Definitions (XSDs), we need algorithms to generate uniformly at random a corpus of XSDs as well as a similarity measure to compare how close the generated XSD resembles the target schema. In this paper, we provide the formal foundation for such a testbed. We adopt similarity measures based on counting the number of common and different trees in the two languages, and we develop the necessary machinery for computing them. We use the formalism of extended DTDs (EDTDs) to represent the unranked regular tree languages. In particular, we obtain an efficient algorithm to count the number of trees up to a certain size in an unambiguous EDTD. The latter class of unambiguous EDTDs encompasses the more familiar classes of single-type, restrained competition and bottom-up deterministic EDTDs. The single-type EDTDs correspond precisely to the core of XML Schema, while the others are strictly more expressive. We also show how constraints on the shape of allowed trees can be incorporated. As we make use of a translation into a well-known formalism for combinatorial specifications, we get for free a sampling procedure to draw members of any unambiguous EDTD. When dropping the restriction to unambiguous EDTDs, i.e. taking the full class of EDTDs into account, we show that the counting problem becomes #P-complete and provide an approximation algorithm. Finally, we discuss uniform generation of We acknowledge the financial support of the Future and

The extension of Regular Expressions (REs) with an interleaving (shuffle) operator has been proposed in many occasions, since it would be crucial to deal with unordered data. However, interleaving badly affects the complexity of basic operations, and, expecially, makes membership NPhard [13], which is unacceptable for most uses of REs. REs form the basis of most XML type languages, such as DTDs and XML Schema types, and XDuce types [16, 11]. In this context, the interleaving operator would be a natural addition to the language of REs, as witnessed by the presence of limited forms of interleaving in XSD (the all group), Relax-NG, and SGML, provided that the NPhardness of membership could be avoided. We present here a restricted class of REs with interleaving and counting which admits a linear membership algorithm, and which is expressive enough to cover the vast majority of real-world XML types. We first present an algorithm for membership of a list of words into a RE with interleaving and counting, based on the translation of the RE into a set of constraints. We generalize the approach in order to check membership of XML trees into a class of EDTDs with interleaving and counting, which models the crucial aspects of DTDs and XSD schemas.