Projects concerned with capturing historical sources are increasingly keen to disseminate their digital results via the World Wide Web. Scholarly resources made available on a website enjoy a number of benefits, not least of which is the potential for wide and convenient access among the academic community. But there are also a number of issues which such projects should bear in mind from the very beginning if they ultimately intend to employ the internet to make their resources available.

The first and most important thing to bear in mind is that while webpages may be a suitable format for disseminating historical resources, they're not at all suitable for either capturing or preserving historical data. HTML is a format designed to fit text and images into a structured layout so that the content of a webpage will appear in a certain way when that webpage is displayed in a browser. Unfortunately for most historians wishing to use digital resources, a web browser is a very limited analytical tool, even taking into account recent advances in web technologies. In terms of preserving historical resources, websites are notoriously ephemeral even when hosted by university departments, and can often fall victim to changes in standards and the way that web browsers interpret HTML.

The solution to both issues, the lack of analytical capabilities and preservation, is to capture and store data in another format initially, and then translate it to HTML later. Building the resource as a database or a text archive, both areas where the AHDS can provide advice, allows the project the flexibility to prepare data for paper, CD-ROM and web-based publication later.

Of course the translation of data from database to webpage isn't a simple matter either. The simplest type of website to build is one that consists entirely of static pages, that is, pages with fixed content. In some situations this is quite appropriate, for example, a critically edited collection of letters from a 19th century politician could be usefully organised on a website according to pre-defined criteria. But often of much more use to scholars is a site where the content is generated dynamically according to the preferences of the user. With the example above, if the collection of letters is large, it may be very helpful if users can locate particular letters by date, correspondent and topic. A dynamic website is a way of enhancing the resource created by a project, but it should be noted that linking a website to an underlying database requires specialist technical knowledge of various kinds of software and hardware, and this is something that a project should consider in its early stages. And of course, whether the website is static or dynamic, it should conform to web accessibility standards.

The other issue that a project should bear in mind before setting up a website is that of rights. When publishing any kind of digitised version of a source on the internet, the project should be as confident as possible that they have the permission of the relevant rights holders to do so. Clearly this is something that should be obtained right at the start of the project. The flipside to this is that some form of control might need to be exerted over who can access the website, as once the resource is launched, it's in the public domain to a greater or lesser extent. Some projects may not feel the need to protect their resource by limiting access to certain groups, but this is a decision that needs to be made.

Mark Merry, History Data Service

Producing a digital copy or surrogate from an original paper copy is a process common to many digitisation projects, but the methodology and technology which underlie this process have remained largely unchanged in the past decade, and fall within two broad areas: Optical Character Recognition (OCR) and re-keying the material.

Optical Character Recognition can be a powerful technique for the rapid digitisation of printed materials. As personal PCs have become more and more powerful, OCR has become a realistic and affordable desktop application. But although it may seem like a simple matter to take a group of pixels and work out what shape they are, it is in fact an extremely intensive computational task, with vast potential for error. As a result, OCR packages will often only be guaranteed to work efficiently with a small subset of character fonts, and will be confused into making errors by defects in the original print quality. It also means that recognition of handwriting from images is still well beyond the capabilities of current personal computers. Modern Personal Digital Assistants (PDAs) which recognise handwritten input can do so because they specify how the user should form their letters with the stylus, and observe the creation of the letter as it takes place. Without these clues, human handwriting is simply too irregular to be analysed accurately using current systems. Handwritten materials, therefore, must still be digitised manually, by keying them in to the computer. OCR techniques are also extremely poor at differentiating the 'layers' of text which make up many documents. For example, an administrative form may contain printed text comprising the form's original field descriptions, a typewritten layer of the information entered on the form, and an official stamp indicating that the form has been processed. Although some OCR packages, such as Omnipage, will attempt to unravel these layers, they are still extremely error-prone. In the case of complex layered documents, it may again be more efficient to transcribe the text by hand.

Another problematic area is that of non-standard character sets. Commercial OCR packages recognise only a handful of modern languages and character variations. If your material falls outside of this subset, you may wish to look into trainable OCR software. Training is commonplace in the world of software speech recognition, where individual voices vary. A user reads a set of words aloud to the program in order to 'train' it to recognise one's voice. Certain OCR projects have adapted this approach. An open source trainable OCR product called Clara is currently being developed. It promises to allow recognition of arbitrary characters, given some training input. Whether the time spent training the software would represent an efficient use of time overall depends very much on the quantity and uniformity of the material to be digitised.

The decision to use OCR must always be taken carefully, after appropriate pilot studies have been undertaken. If the work proves to be too difficult or too time-consuming, one option would be to contract it out. Whatever software you use, ensure that proper proof reading is done at all stages. Print that appears uniform to the human eye can vary greatly when it is analysed by the OCR algorithms. And finally, retain all the scanned images where possible. They are a resource in themselves, and there are many reasons why these images may prove useful in the future, for your own project or for other researchers. If nothing else, they can be extremely useful when testing new OCR software!

More information on OCR can be found in the Guide to Good Practice, Creating and Documenting Electronic Texts, and also on the website of the Higher Education Digitisation Service,

Alan Morrison, Oxford Text Archive

While the AHDS is happy to receive telephone calls and emails to discuss aspects of a digitisation project, many of the principals that those undertaking digitisation need to be aware of are laid down in our series of Guides to Good Practice, available from . The AHDS has recently published electronic versions of three new guides.

The first, CAD: A Guide to Good Practice written by Harrison Eiteljorg II, Kate Fernie, Jeremy Huggett and Damian Robinson and produced by the Archaeology Data Service (ADS), is based on the premise that many organisations create and hold digital project archives involving computer aided design (CAD) files and 3-D CAD models, digital objects that one cannot reproduce on paper. For this reason the CAD Guide to Good Practice is aimed at the creators of CAD files, primarily those produced in research. The guide provides a basic description of CAD software; a discussion of the use of CAD in a variety of situations; descriptions of data acquisition and capture methods; and information about archival practices. As well as providing a source of useful generic information, the guide emphasises the processes of long-term preservation, archiving and effective data re-use. As a result, the importance of adhering to recognised standards and documenting essential pieces of information about a given resource are recurring themes. The guide will be of use not only to those in Higher Education, but also national heritage agencies and local authorities; and those managing museums, National Monuments Records and Sites and Monuments Records. CAD is a foundation on which many mapping applications are built so this guide will also be of use to any scholars working with GIS.

The second guide, A Place in History: a Guide to using GIS in Historical Research by Ian Gregory, has been commissioned by the History Data Service (HDS). The guide is intended for historians who want to use Geographical Information Systems (GIS), describing how to create GIS databases and how to use GIS for historical research. The author defines GIS; evaluates the way GIS models the world; describes how to get data into a GIS; demonstrates the basic operations that GIS offers to explore a database; reviews how time is handled; explains how GIS can be used for simple mapping and more advanced forms of visualisation; discusses quantitative data analysis within GIS; illustrates the use of GIS for qualitative analysis; and provides help on how to document and preserve GIS datasets. It also includes case studies from a variety of historical projects and an extensive reading list of GIS texts relevant to historians. A minimum of jargon has been used throughout, and no prior knowledge of GIS has been assumed. The aim is to highlight the main themes in using GIS and provide references that allow the reader to follow them up in more detail. There is an extensive literature on GIS in general but in terms of historical research this literature is limited and widely scattered. This guide attempts to bring this literature together to illustrate how historians have used GIS.

The final guide is Creating Digital Audio Resources, authored by Nick Fells, Pauline Donachy and Catherine Owen, and intended as a basic 'how to' guide for those wishing to use audio materials in the creation of digital resources. It deals with such issues as copyright, choosing equipment, playing audio media, delivery of audio to users, data management and preservation. While not intended as an exhaustive guide and while it will almost certainly be necessary for resource creators to consult additional, more specialised sources, the guide provides a vital opening to the issues in developing digital sound.

Catherine Hardman, Archaeology Data Service

To complement the AHDS' core services of collecting digital resources and advising on related issues, a significant research and development programme of projects exists across all its service providers. The primary aim of such projects is often to produce particular end products, such as improved systems or resources for research, learning and teaching. However, the undertaking of such projects by AHDS service providers also provides additional benefits to the communities we serve. One such case in point is the Visual Arts Data Service’s (VADS) JISC-funded Learning and Teaching project: PICTIVA - Promoting Image Collections for Learning and Teaching in the Visual Arts, the fruits of which are now becoming available.

The project has produced tools and resources to enhance the delivery of VADS visual arts image collections and these are going through final evaluation stages before being made available via the VADS website. New features on VADS image catalogue, produced through the project will allow users to:

• Save, manage, annotate and disseminate personal image-sets (Lightbox)
• Find visually similar images by colour, contrast and shape ('Content Based Image Retrieval' - CBIR)
• Engage with exemplar learning resources based on VADS collections, to illustrate the applicability of digital images to learning and teaching.

These have been developed to specifically benefit learning and teaching in the visual arts. However, a number of broader benefits to other activities and communities will result.

An example of this is in the applicability of the tools beyond the learning and teaching community to those engaged in research. The ability to save images into personally managed image-sets will allow researchers to gather, retain, annotate and distribute images according to particular interests and needs. This will also increase the potential for VADS systems to be utilised as a communication tool, rather than solely as an informational one, with the potential to stimulate debate and research.

The research community will also benefit from VADS’ development of its online delivery systems to cater for the production and publication of learning materials. The same technology can be readily applied to produce other types of publications based around material deposited with VADS, such as scholarly papers, and to facilitate the analysis of individual images, sets of resources or whole collections.

Similarly, the Content Based Image Retrieval tool will increase access methods for researchers as well as learners and teachers, enabling all users to better locate material of relevance to their interests, without having to rely on textual descriptions. Additionally, the feature will also provide an exemplary implementation of this cutting edge technology for researchers in the field to analyse. The visual search has been implemented by the Institute for Image Data Research, University of Northumbria and feedback will inform their continuing work in this area, as well as others doing similar projects.

This indicates another broad benefit of PICTIVA and project work generally. End products and findings can often be applied across subject boundaries and also assist interdisciplinary activities. This will principally be across other arts and humanities subjects, through the application of the tools to locate and utilise materials for various subject-specific needs, but can also extend to the technologies themselves being the subject of enquiry for a broader range of disciplines.

The final area of broad benefits of the project results from the investigation of new technologies and methodologies that VADS has undertaken in developing its systems. The PICTIVA project has seen VADS engaging in Human Computer Interfacing issues, such as good practice for navigating online resources intended for direct-user interaction; Learning Object Metadata for appropriate resource discovery and project management issues more generally. The lessons VADS have learnt add to its knowledge base and this expertise can be passed on through VADS advisory activities.

To summarise, much is learnt through project work that can inform an organisation's own future practices. But more importantly for an organisation such as the AHDS, expertise gained can be passed on to a range of users and interested communities.

Phill Purdy, Visual Arts Data Service

The project's goals are linked to the development of a future joint Humbul / AHDS Art & Humanities portal that may use powerful subject browsing and advanced searching across distributed digital collections. In order to meet these goals, the controlled vocabulary will be backed by a classification structure, allowing:

• Structured and flexible browsing
• Unlimited combination of concepts while indexing (e.g. pre-coordination)
• Expression of the relationships/roles of different concepts in compound and complex index entries
• Links between concepts in the classification structure and the natural language terms that are expected to be used in retrieval

Main foci of the Project

In order to make full use of facet analysis features, a vocabulary has to be adequately modelled and implemented for testing. Therefore, the project has three equally important foci:

1) Building a knowledge structure

Basic classification vocabulary (compiled from existing faceted classification schemes):

• A broad classification structure, universal in scope
• A faceted knowledge structure suitable for all the humanities fields
• A set of generally applicable facets, e.g. Place, Time, Materials, Person, Form, Languages, Properties, Processes

Controlled vocabulary for the humanities:

• A newly developed and full faceted subject vocabulary; target areas for the prototype will be religious studies, the visual arts and history

2) Building a data model to hold the knowledge structure and vocabulary

3) Model implementation, and prototype testing and evaluation

FATKS will provide a subject browsing structure for the humanities portal and underlying local collections. The structure is envisaged as a general knowledge representation 'umbrella' that may be later linked or mapped to vocabularies particular to each of the source databases encompassed by the portal, or to any other subject organisation tools relevant for the humanities (e.g, Dewey Decimal Classification or the Art and Architecture Thesaurus).

For more information please visit the FATKS website at

Aida Slavic, University College London