The paper describes a collaborative project, funded by the UK Arts and Humanities Research Board, between the School of Library, Archive & Information Studies, University College London and two major digital resource gateways, Arts & Humanities Data Service, and the Humbul Humanities Hub. AHDS (http://www.ahds.ac.uk) and Humbul (http://www.humbul.ac.uk) are on-going government funded projects for the identification, evaluation and organization of quality digital resources in the humanities, primarily for the use of the higher education community. AHDS' remit includes visual and performing arts and archaeology in addition to traditional humanities disciplines; Humbul covers a slightly narrower humanities field including history, archaeology, literature, theology and philosophy. The two are developing a single humanities portal (http://www.portal.ac.uk) which will become operational in 2002. The new portal will draw in resources from the wider Web in addition to the managed material already available. An important consideration is the choice of a tool to manage the subject content of the new site. A digital library on the AHDS model has much in common with the conventional library in terms of structuring the semantic content of the resource; it may benefit from the knowledge organization theory that has been developed over the last fifty years within the library sector for the creation of tools for vocabulary management and semantic organization of document content. Systems such as faceted classifications, structured subject headings, thesauri, and other controlled vocabularies provide a scientifically based approach to the analysis of 'document' content, and to the creation of indexes, descriptors, visible taxonomies and hierarchies, as well as linear ordering schemes (i.e. rules for filing order and sequencing) for the physical management of materials with respect to intellectual content. These have been tested over managed bibliographic databases as well as print-based materials, and the theory is at a high level of sophistication. Existing means of subject organization at AHDS and Humbul are the Library of Congress Subject Headings (LCSH) and the Dewey Decimal Classification (DDC), both designed for organization of print-based material in a traditional library. While these offer management advantages (e.g. an established system with institutional support, regular maintenance and revision, and centralised bibliographic services) they are not particularly useful within a digital environment. They display little sophistication in the structure, cannot handle complex objects well, and can do little to expose the complex interrelationships and multidimensional links within the structure of the digital collection. The new humanities portal requires a system that performs several functions;

• accurate description, for retrieval purposes, of complex digital documents/objects with a range of attributes, both of intellectual content and format
• provision of a systematic structure for the organization of the front-end in a directory format, using hypertext techniques to expose deeper layers of the network
• generation of structured subject headings for specific objects
• manipulation of these to create browsable alphabetical subject indexes
• capability of conversion to a thesaural structure to provide a controlled vocabulary of keywords and concepts.

Ideally, the system should also display;

• potential for multiple access points to the structure to enable resource discovery by various routes or search strategies;
• potential for incorporation into search software as a device in negotiating the wider Web.

The School of Library, Archive & Information Studies (SLAIS) at University College London has a particularly strong history in education and research in classification and indexing. It is one of only a few British schools offering teaching in this area, and its staff are actively involved in the management of several systems of bibliographic classification, and research into the development and use of faceted schemes. We are investigating a structure of this kind for the organization of the new humanities portal. Classifications built on the facet analytical model provide effective tools for analysing and organizing documents on the basis of their subject content, and consequently for retrieving those documents from a managed store. They work on different principles from older enumerative schemes such as Dewey and the Library of Congress classifications which simply provide long lists, or enumerations, of classes for the accommodation of documents. Facet analysis was conceived by S. R. Ranganathan, a mathematician by training, and a student at SLAIS in the 1920s. He proposed a system for the description and organization of documents with complex subject content, based on identification and analysis of constituent parts of the subject, rather than by creation of lists or enumerations of compound classes into which specific documents must be fitted. Documents were analysed, the content encoded, and the codes synthesised into an appropriate classmark which was used for filing and which was expressive of the subject content. Ranganathan's system was ground-breaking, but relatively unsophisticated. It continues to be developed in India, but is virtually unused outside that country. In the UK the Classification Research Group, formed in the 1950s, further developed facet analytical theory. The internal logic of a faceted system of the CRG type is based on rigorous analysis of the terminology of a subject, whereby terms are sorted into standard sets of functional categories. Within these categories a range of semantic relations are acknowledged, and problems of vocabulary control (such as synonymy, partial synonymy and variations in word forms) are addressed. A sophisticated system syntax provides for arrangement and combination of terms both intra- and inter-category, and for the management of syntactic relations. This improves performance in the accommodation of complex subjects, the predictability of location, and in the effectiveness of retrieval. A faceted classification is, in its simplest form, a structured set of simple terms or concepts with rules for the combination of these into compound concepts such as occur in the content of documents. These compounds are placed precisely in the base structure by the application of the system syntax. When these classes are populated by the 'real' subjects of documents (or other objects with semantic content) a more complex structure grows in accordance with the internal logic of the system. A faceted classification, when applied to a large collection of documents, can generate a very complex knowledge structure of n-dimensionality and great logical regularity, with deep levels of hierarchy. The resultant structure can be utilised in a number of ways; as an ordering device, as a source of index terms and subject headings, and can also be converted to a thesaurus. Hypertext can be utilised to expand the levels of hierarchy, or to make links between distributed elements. An example of a small classification for religion demonstrates how the structure can be applied:
Judaism (Form subdivisions) Bibliography of Judaism Encyclopaedia of Judaism (Place subdivisions) Judaism in Europe (Period subdivisions) Judaism in the Middle Ages Judaism in the Nineteenth Century Judaism in Nineteenth century Europe (Philosophy and theory of religion) Religious philosophy of Judaism (Sacred texts) Hebrew Bible Mediaeval Hebrew Bible (Worship) Jewish festivals (Organization of the religion) Jewish religious law (Sacred texts) The Hebrew Bible in Jewish religious law This can be represented in the form of subject headings as:
Judaism Judaism - Bibliography Judaism - Encyclopaedias Judaism - Europe Judaism - Middle ages Judaism - Nineteenth century Judaism - Nineteenth century - Europe Judaism - Religious philosophy Judaism - Bible Judaism - Bible - Middle Ages Judaism - Festivals Judaism - Religious law Judaism - Religious law - Bible These can be left in this order, to represent the systematic structure, or they can be alphabetized:
Judaism Judaism - Bible Judaism - Bible - Middle Ages Judaism - Bibliography Judaism - Encyclopaedias Judaism - Europe Judaism - Festivals Judaism - Middle ages Judaism - Nineteenth century Judaism - Nineteenth century - Europe Judaism - Religious law Judaism - Religious law - Bible Judaism - Religious philosophy From a small base vocabulary of 30-40 terms like this one, hundreds of multi-term subject headings can be generated. The subject headings can be inverted to form a browsable index in which distributed relatives are collocated. Bible - Judaism Bible - Religious law - Judaism Bibliographies - Judaism Encyclopaedias - Judaism Europe - Judaism Europe - Nineteenth century - Judaism Festivals - Judaism Judaism Middle Ages - Bible - Judaism Middle Ages - Judaism Nineteenth century - Judaism Religious law - Judaism Religious philosophy - Judaism The regularity of the system and its rules of syntax suggests that much of the routine work of managing documents could be carried out automatically, once the initial intellectual analysis has been made. In a testbed implementation for the research, AHDS and Humbul are applying the knowledge structure to the Portal's planned metadata repository for all the digital objects in their collection; it is likely that XML will prove to be the best tool for the implementation of the structure. They will also experiment with its use in cross-disciplinary browsing and retrieval of digital resources which are held elsewhere.