Object-oriented modeling of multimedia documents

Abstract

This paper describes an object-oriented model for paper-based multimedia documents such as a textbook with embedded graphics. This model is the first step towards building a manageable authoring system for the Web, in which documents can be easily built, extended, truncated, reordered, assembled and disassembled on a component basis, and the document components can be reused. The model will also make accessible properties which might be significant or important to the user, especially in searching or classifying documents, such as the document title and author. The model design is explained, and the class hierarchy for the model is presented.

Keywords

Object-orientation; Modeling; Multimedia; Documents

1. Introduction

This paper presents a manageable multimedia document model, using the proven object-oriented technology. Documents can be built, extended, truncated, reordered, assembled and disassembled on a component basis, and document components can be reused. The model also makes accessible properties that might be significant or important to the user, especially in searching or classifying documents, such as the document title and author.

The design criteria of the model are completeness, compactness, and simplicity. The model attempts to cover all types of existing paper-based documents. The number of classes, their attributes and methods are kept to the minimum. Generic classes are used wherever possible, distinct document types may be merged into one class, and convenience functions and attributes are ideally kept to none. The system derived from this model is meant to be an open system, ready for use by the public. Implementations of this specification must be easy for general users to carry out. This model is designed for electronic publishing, but is based purely on traditional paper-based publishing. Factors relating to hyperlinks and distributed environments are not covered. It is to be used as the base for integrating traditional publishing into electronic publishing.

2. The Multimedia Document Model (MDM)

A document object represents the entirety of a publication, and is constructed by the basic building block of structure. Instances of the document class correspond to the loosest form of publication. It is designed to cover all types of documents in the traditional document system, be it an unbounded draft, letter or bounded journal. The document object can be composed of only structure objects, or some structure objects and some media objects (structured documents), or just media objects (unstructured documents). The only two attributes that require non-null values are contents, as we do not allow empty documents, and title, which is used to enforce that its content is semantically complete, ie. it must not contain just fragments of text or graphics which are not comprehensible to human readers. The title also serves as a mnemonic identifier to human readers, but does not have a significant meaning to the system, and is not required to be unique in any domain.

A structure object, inheriting the aggregation relationship to itself from document, can in turn contain itself recursively. A document is therefore composed of a recursive chain of structure objects, which can be assembled, disassembled, and reordered, allowing for the whole document to be modified, extended, or truncated on a component basis, without losing the coherence of its overall structure. Structure is the generic class for all structural components of a document, such as volumes, parts, chapters, and sections. Our model therefore allows for organisation such as a volume containing several parts, a part containing several chapters, a chapter containing several sections, etc., commonly found in paper-based documents. Since all these structural components share the same properties and behaviour, it is most appropriate to have one single representative class. This design makes the model simpler and more compact. It also gives authors more flexibility when defining their own document since the model does not differentiate between various structural components.

The model also makes available presentation components, by defining them as children classes of structure. A structure object can thus be presented in a number of different formats: text chunk, table, frame, list, reference entry, and article. Text chunk is a sequence of one or more ASCII characters. It is essentially a section of text, with a heading and one or more paragraphs. Table, frame, list, reference entry and article, are all defined in common sense as have been used in the traditional publishing industry. Text chunk is the default presentation for a structure or document object. That is, with reference to Fig. 1, when a structure or document object is resolved to one or more media objects (see Section 2.2 for explanations about document resolution), it will assume a text chunk presentation. If an object of a descendant class of structure is resolved, it will take other presentations according to the class to which it belongs.

Media objects hold the real contents that make up the whole system. Their content is unstructured, raw materials that will be used to fill in the document objects or their structural or presentational components. The actual data type of the content will be defined in the offspring classes of media. In the current version of MDM, it can be either text or graphics. This content can be semantically incomplete, ie. not meaningful to human readers. For example, it can be a fragment of text, which is a part of chapter of a book. This semantic incompleteness is reflected in the object design by the fact that no title or heading is required for an object of the media branch. Offspring objects of the media class, representing different media types, can therefore be inserted at arbitrary places in a document object. This concept is applicable to all kinds of media types, although at the moment it is meaningful to the text object only. Consider a chapter in a textbook with an image inserted halfway. If the chapter is defined as one document object, then it must consist of three media objects of two broken text pieces and one graphics (see Fig. 2).

Publication is the generic class for all types of published materials. In real life it generally corresponds to any publication, normally bounded, that can be identified by an international identification number, usually ISBN (International Standard Book Number) or ISSN (International Standard Serials Number). There are more specific classes inherited from Publication, designed to handle other specific features of different document types. All of these classes are designed based on the existing paper-based document system as found in most libraries nowadays. For compactness and simplicity purposes, several library categories are merged into one single class in our model.

Book includes such publications as textbooks, storybooks, manuals (non-references), the Bible, and picture books. A book object normally has only one authorship (with one author and possibly co-authors, but all cooperating in one work) and typically contains items that are closely related to each other. For example, there are normally references to other sections from one section in a textbook. A book is typically a non-collective, non-recursive entity. Some exceptions are the Bible, a book of short stories, a condensed book, etc, which can contain other books.

Serial includes magazines, newspapers, journals, conference proceedings, research reports, bulletins, newsletters, etc. They are published periodically and all have some method for sequential ordering, either by numbers or dates, and their contents are collectively authored. They typically contain a collection of independent writings, each of which is complete and comprehensible without the need to refer to or understand other items in the collection (although they can use other references outside the collection as references). Some examples are articles, technical papers, reports, lecture notes, advertisements, and meeting notices. A collection of serials can be bound into a volume, identified by a volume number.

Reference covers all reference materials, such as dictionaries, handbooks, calendars of events, directories, and reference manuals. They normally contain many reference entries, which can be grouped into sections. Each entry usually consists of a reference term and its explanation or specification. The reference class has a special method, sortEntry().

The design of our model allows the mapping of all existing document types. The book, serial and reference classes cover most library printed materials. The publication and document classes account for any other types, both published and unpublished. Each offspring class of publication is actually a merge of many library categories, and new classes are created only if there are important properties or methods that need to be distinguished between them, such as the issue number for serial objects, or the sortEntry() method for reference objects. The model is thus both complete and compact, since it covers all document types while the number of classes are kept to minimal. Simplicity is achieved by compactness, and also by the fact that the model design is based on the familiar and well-developed traditional paper-based system, with many defacto standards for structural and presentational components. This provides a solid base, making the model robust to changes and simple to use.

2.2. Document resolution
Objects of the document class (and all objects of its offspring classes) are composed of text and graphics objects, which belong to the offspring of the media class. Every object of the document branch must be resolved to one or more objects of the media branch. In other word, the document object can be considered the virtual container of the system contents, holding only references to media objects, which keep the real contents of the system. The need for separating between media and document is to support the three views of the document (see Section 2.3), making both media and structural components accessible. It also allows for media objects, which are pieces of text or graphics to be reused in more than one document. This feature shall reveal an even more important role in the distributed environment, where virtual contents can be distributed.

2.3. Three views of the MDM
In the MDM, a document is viewed in three dimensions: structure, media and presentation.

In the structure view the document is broken up into building blocks, such as volumes, parts, chapters, sections, and paragraphs. The decision of the level of granulity is the result of a tradeoff of two major influencing factors: the simplicity and compactness of the whole system, and the dynamics of the component under consideration. In Fig. 1, the structure view is represented by the generic class structure. This class is defined to cover up to the section level (see Section 2.1). Thus in normal terms we can say that the basic building block of the MDM is section. (Note that in our model the term section has been changed to the generic term structure). This view makes the structural components of the document accessible, allowing for the document to be managed structurally on a component basis (see Section 2.1).

The media dimension accounts for different media types that compose a document, such as text, graphics, sound, and video. Only text and graphics are considered in this system, since it is restricted to the current paper-based document system. Other media types however, can be easily added in the next version of MDM, because of the extensibility characteristic of object-oriented systems. In Fig. 1, the media view is represented by the class media and its offspring classes. This dimension is a necessity in the multimedia environment. As media components are made independent entities, separate from structure components, each can have its own properties and behaviour specific to its media type, and their layout in relation to other objects can be controlled. For example the alignment of a graphics object can be specified relative to other text objects surrounding it with the hAlign and vAlign attributes. The separation of media from document also allows for media objects to be reused in more than one document. This is especially useful in the World Wide Web environment where graphics such as a company logo and a navigation bar, can be included for every Web page. This sense of separation is stronger in the Web, where objects of different media types must be saved in separate files and treated independently, for example, text is stored in text files, and graphics in GIF files.

The presentation view looks at various ways in which information is presented to human readers. Presentation has only limited support in the current version of MDM. Other format and presentation issues are intentionally left out for the next version, or for use of other mechanisms such as the Cascading Style Sheet (CSS) [CSS97]. All the offspring classes of the structure class, table, frame, reference entry, and article, represent the presentation view in our model. The presentation view makes it possible to present structural components in different formats.

In summary, the three dimensions together make the document components accessible in various ways, allowing for the document to be assembled or disassembled structurally, its layout suitably controlled for components of different media types, and various formats can be presented.

3. Related work

Other related works include the metadata initiatives such as the Dublin Core (DC) [W3C98]. DC and our work are similar in an interesting aspect in that they both use data to describe data. A document object holding information about itself is essentially a set of metadata. DC and MDM however, come from two completely different aspects of software engineering, with two different types of model. MDM also has a broader goal for the management and maintenance of the whole system, not just for resource discovery and cataloging as does the DC.

4. Conclusion

References
[RUM91] Rumbaugh, J., Object-Oriented Modeling and Design, Prentice Hall, 1991.

[CSS97] Cascading Stylesheet (CSS2) Specification, W3C Working Draft, 1997, http://www.w3.org/Style/

[GEL97] Gellersen, H.-W. et al., WebComposition: an object-oriented support system for the Web engineering lifecycle, in: Proc. WWW6 International Conference, 1997, http://www6.nttlabs.com/HyperNews/get/PAPER232.html

[ING95] Ingham, D.B. et al., Bringing object oriented technology to the Web, in: Proc. WWW4 International Conference, 1995, http://w3objects.ncl.ac.uk/pubs/bootw/; also at http://www.w3.org/pub/Conferences/WWW4/Papers2/141/

[ING97] Ingham, D.B. et al., Supporting highly manageable Web services, in: Proc. WWW6 International Conference, 1997, http://w3objects.ncl.ac.uk/pubs/shmws/; also at http://proceedings.www6conf.org/HyperNews/get/PAPER27.html

[NEX97] WebObjects, http://www.apple.com/webobjects/

[W3C98] W3 Consortium, http://www.w3.org/