Abstract: "The ISIS European XML/EDI Pilot Project has used Best-Practice Guidelines to disseminate its findings to the EDI and XML communities. This paper seeks to explain why the project team chose to create a set of guidelines rather than simply develop finished solutions to specific problems. It also summarizes the recommendations for further standardization work which were the raison d'être of the project."

"The work done on converting UML models into DTDs suitable for the interchange of healthcare messages between healthcare professionals (as opposed to just between their computer systems) was recorded in a paper entitled Mapping from UML Generalized Message Descriptions to XML DTDs. Automatable rules were defined for the conversion of each type of construct permitted in a UML model, together with a set of semi-automatable rules for conversion of UML object names into XML element or attribute names. A program was developed to manage this process, and a suite of related healthcare messages was developed from the UML model previously developed as part of the European prestandard for Electronic Healthcare Record Communication (prENV13606). The rules used while converting the selected MIGs for transportation messages were recorded in a paper entitled Mapping Existing EDIFACT MIGs to XML DTDs. A set of 16 basic rules was defined, with explanations as to why they need to be applied, and an example of how each rule was applied within the DTDs developed for the project. Once both sets of rules were drafted the project team studied the relationships between the two sets of rules to ascertain what features they had in common, and why they produced differing results. From this study they were able to identify conclusions that were of general applicability, which were recorded in a paper entitled Best Practices for Creating XML/EDI DTDs. This highlighted the need to simplify messages before converting them for use in an XML/EDI environment, the advantages of formally modeling the message to ensure that all relationships between elements were fully recorded, and the need to have consistent rules for creating meaningful names for XML markup. The study also highlighted the advantages to be gained by using default and fixed attribute values to record information that was not likely to change from message to message. When it came to documenting best practices for the use of XSL Transformations and other members of the XML family of standards it was considered to be advisable to target these practices at different processes. For this reason the project team produced reports covering the role XSL can play in creating data capture user interfaces, in validating XML messages beyond the limits provided by XML DTDs (or XML schemas), in presenting data on web browsers and mobile phones, and in managing the processing of transactions.[...] The project team's recommendations highlight a range of other actions required, including a need to extend testing into other application areas, the need to test the next generation of W3C standards as they are finalized, and the need to actively promote the harmonization of semantic definitions. The need for semantic repositories that could be queried using a standard, web-compatible, method which would allow either a description of relevant elements/attributes to be returned, or their formal definitions, was also highlighted. The project team has also prepared a short document on Using XML for Electronic Data Interchange, which seeks to explain the basic concepts used in both XML and EDI in terms that can be understood by small and medium sized enterprises (SMEs) who do not have a background in IT. Unless we can convince such companies to adopt XML-based techniques for EDI we will not be able to fully realize the potential of XML/EDI within Europe, where more than 90% of firms are SMEs. The goal of the European Commission in setting up a single market within Europe will be greatly enhanced if widespread use of electronic data dissemination can be introduced, both between companies and between them and the government bodies that serve them. For this reason the European Commission has been, and will continue to be, a champion of XML/EDI."

Abstract: "Semi-structured data, including but not limited to structured documents, has specific characteristics and is used in ways different to tabular data. SGML and XML are widely used to represent information of this type. The demands on systems that manage semi-structured data vary from those on traditional relational systems. This paper reviews the nature and characteristics of semi-structured data, and the functional needs of those applications, including query requirements, document description, manipulation, and document management needs. It examines alternative physical models for semi-structured data, and evaluates and compares alternative system architectures."

[Conclusion:] "Structured documents are a form of semi-structured data. The characteristics of structured document data are quite distinct from those of, say, relational data. Documents are hierarchical, textual, variable in size, variable in structure, and self-describing. The uses of document data also impose specific requirements: queries must be able to target or be constrained by documents, document elements, metadata, content, structure, and hyperlinks; transformation of document structure is important; a data definition may not always be available, or may be relatively general; document management needs may require support for component management, versioning, document and component-level locking and access control. Several main approaches exist for implementing document content management systems, including on top of an RDBMS, as an OODBMS application, or as a specialized, structure-aware system. Basing a content management system on an RDBMS or OODBMS creates problems resulting from the mismatch between document data and those approaches' data models. The variability, inconsistency, and potential unavailability of concrete data definitions for semi-structured data such as documents complicates data modelling and representation in RDBMS and OODBMS based systems. In the relational approach, a significant overhead results from fragmentation of document data across multiple relations. Operations such as document transformation must be carried out externally to the RDBMS engine. In contrast, a structure-aware system benefits from directly representing document data by avoiding fragmentation; being able to use documents' own schemas as a data definition and for validation; being able to integrate content, structure, and metadata query processing; and supporting document transformation internally. Document management functionality may also be affected by the choice of approach. Under a relational approach, all document management must be performed by a separate application layer that reassembles documents or document components as required. Under an object-oriented approach, object level locking and access control may be directly handled internally to the OODBMS; however, component management and versioning support can not. A structure-aware approach may be able to internalize all document management-related functionality, including access control, component reuse, versioning, and version indexing. Low cohesion between system components in architectures based on RDBMS or OODBMS engines places overheads on content management system implementations; these are avoided or minimized in the simpler structure-aware architecture where there can be tight coupling between an internal XML-processing engine and other system components."

Abstract: "The new ISO standard ISO/IEC 13250 Topic Maps defines a model and architecture for the semantic structuring of link networks. Dubbed the 'GPS of the information universe,' topic maps will become the solution for organizing and navigating large and continuously growing information pools, and provide a 'bridge' between the domains of knowledge representation and information management. This paper presents several technical issues of which are of great interest when applying topic maps to real world applications. The main focus of the paper is the introduction of 'topic map templates' -- a semi-official term coined by the standards' committee for a concept that the author argues is a necessary but as yet unstandardized addition to the basic model. Furthermore: association taxonomies, class hierarchies, and consistency constraints of topic maps are presented and discussed."

[Conclusion:] "The new topic map standard ISO/IEC 13250 defines a model and architecture for the semantic structuring of link networks. It can be seen as a base technology for modeling knowledge structures. The standards working group defined topic maps in such a way that a limited but implementable set of core concepts express the necessary semantics. The STEP Group has investigated how topic maps can be applied to reference works and uncovered some concepts which are not made explicit in the standard: (1) ability to separate the declarative part from the 'real' map, (2) predefined association types and association type properties, (3) class hierarchies for types, and (4) consistency constraints as input to map validation. The paper has explained these concepts and presented meaningful solutions. First experiences have shown that the part of a topic map made up by all topics used as themes and types by other 'objects' in the map should be clustered somehow. For this purpose the term topic map template was coined by the ISO working group. Templates can be used as starting points for new maps or can be used by reference in order to provide all the themes and types the map needs. Standardizing topic map templates will offer base topic maps for specific application areas and could form the basis of semantic application profiles. We looked at related academic fields like mathematics, linguistics, and philosophy to get some substantial input about relations. The results are a list of association type properties which give important hints to the topic map software and a list of basic association types which could act as built-in superclasses. The introduction of the superclass-subclass relationship was the logical consequence. Another technical issue covered by the paper is the validation problem. Topic maps might become rather big with millions of topics, occurrences, and associations. Manual consistency checking will be impossible. All the previously defined concepts open the possibility for sophisticated rule-based validation of topic maps. The proposed consistency constraints are those rules which declare the semantics not expressible with DTDs and which control the validation process. A couple of examples proved that standardizing the missing concepts as predefined topic map templates will help both the topic map developer and the topic map user. The improvements were presented on a level that they can be used as input to the ISO working group for further discussions."

Abstract: "Users are increasingly accessing the Internet from information appliances such as PDAs, cell phones, and set-top boxes. Since these devices do not have the same rendering capabilities (display size, color depth, screen resolution, etc.) as desktop computers, it is necessary for Web contents to be adapted, or transcoded, for proper presentation on a variety of client devices. In this paper, we present a framework of external annotation, in which content adaptation hints are associated with Web documents as separate XML files. On the basis of the framework, we introduce an annotation vocabulary for transcoding, which is employed for the adaptation of HTML documents to small screen devices. We then explain a prototype annotation tool that allows visual authoring for creating external annotations in accordance with the vocabulary for transcoding. Finally, we compare our approach with related work."

"[...] The basic ideas behind this annotation framework are twofold. One is that new elements and/or attributes should not be introduced into a document type definition of subject documents. The other is that annotations may need to be created for arbitrary parts of the subject documents... External annotation files contain hints associated with elements in an original document. RDF is used as the syntax of annotation files. The RDF data model defines a simple model for describing relations among resources in terms of named properties and values. The RDF data model, however, does not define concrete semantics in any application domains. Although the annotation vocabulary introduced in this paper is to characterize Web resources, namely HTML/XML documents, the vocabulary does not specify a tag set for describing individual Web resources in itself. The role of the proposed vocabulary is to constrain the possibilities of content alternation, decomposition, and combination. In addition, XPath and XPointer are used for associating annotated portions of a document with their annotations. An annotation file, which is an XML document, therefore contains a set of RDF descriptions for annotating a subject HTML file. [...] The external annotation framework presented in this paper is applicable not only to transcoding for Web-enabled personal devices, but also to other cases in which content adaptation is desirable. For example, when HTML documents are translated into multiple target languages by means of a machine translation engine, linguistic annotations for proper nouns would be helpful for improving the translation accuracy. In other situations, content adaptation may be needed, so that user-side constraints can be met or alleviated. For example, text contents should be transcoded into audio content for a user who is driving a car. As an example of content adaptation engines, we are currently developing a run-time module of page fragmentation for small screen devices. It is then needed to consider the importance and role of elements in a subject document, so that the document can be split into smaller pages appropriately. Existing HTML documents are, however, not always designed taking account of this possibility. External annotation can then be a useful approach to providing additional meta-information for page fragmentation. We are investigating an authoring environment that allows authoring-time content adaptation with collaboration between the annotation tool and a run-time module for the page fragmentation.

Abstract: "We present a novel, yet simple, technique for the specification of context in structured documents that we call caterpillar expressions. Although we are primarily applying this technique in the specification of context-dependent style sheets for HTML, SGML and XML documents, it can also be used for query specification for structured documents, as we shall demonstrate, and for the specification of computer program transformations. From a conceptual point of view, structured documents are trees, and one of the oldest and best-established techniques to process trees and, hence, structured documents are tree automata. We present a number of theoretical results that allow us to compare the expressive power of tree automata. In particular, we demonstrate that each caterpillar expression describes a regular tree language that is, hence, recognizable by a tree automaton. Finally, we employ caterpillar expressions for tree pattern matching. We demonstrate that caterpillar automata are able to solve tree-pattern-matching problems for some, but not all, types of tree inclusion that Kilpeläinen investigated in his Ph.D. thesis. [Tree Matching Problems and Applications to Structured Text Databases.] In simulating tree pattern matching with caterpillar automata, we reprove some of Kilpeläinen's results in a uniform framework."

The contextual technique we introduce is also applicable to the compilation of computer programs, but has little appeal since compiler designers and writers do not usually allow users to modify a compiler according to new context dependencies. Our techniques may, however, be used when developing code optimizers or other program transformation tools since, in both cases, there may be a number of individuals collaborating on the development. Once we have isolated the specification of contexts from the more general specification of style sheets, we are able to provide naive users with better support for this aspect of style specification. Indeed, it also frees us to consider different techniques (different from attribution, for example) for context specification. Since regular expressions are understood by many people who are not programmers per se, and they are a simple specification technique, we decided to use them for context specification. There is a body of somewhat related work, which we discuss below ('Caterpillars and context'), in which a similar decision was made. We make the well-accepted assumption that a set of similar documents are modeled by syntax trees or abstract syntax trees of a given grammar (an SGML document grammar, an XML document grammar or HTML) that generates the set of all such documents. From now on we will no longer mention SGML and HTML but restrict ourselves to XML and XML document grammars. Indeed, for this paper it is irrelevant which specific grammar mechanism is used to define classes of documents. We introduce and motivate ('Caterpillars and context' and 'Evaluating caterpillar expressions') the notions of caterpillars and context and establish a basic complexity result for the evaluation of caterpillar automata on document trees. In another section ('Caterpillar-regular and regular tree languages'), we investigate the expressive power of caterpillar automata in comparison with tree automata. In particular, we demonstrate that each caterpillar expression describes a regular tree language that is, hence, recognizable by a tree automaton. Finally, in the following section ('Caterpillars and tree pattern matching'), we demonstrate that caterpillar automata can be used to solve tree-pattern-matching problems... We have mentioned our interest in document classes that are constrained by some type of grammars, for example by XML DTDs. The results in this paper hold for the classes of all documents that are given as trees over a fixed alphabet. Applications, particularly applications based on document databases, often come with non-trivial document grammars. It is therefore pertinent to generalize our results to document classes defined by non-trivial document grammars and to address the open questions in this light. Children, and even adults, were able to draw complex figures very quickly in the Logo language using the turtle as metaphor and guide. We hope that our use of caterpillars will garner a similar response from graphics designers."