METS as an Intermediary Schema for a Digital Library of Complex Scientific Multimedia Richard Gartner INFORMATION TECHNOLOGY AND LIBRARIES | SEPTEMBER 2012 24 ABSTRACT The use of the Metadata Encoding and Transmission Standard (METS) schema as a mechanism for delivering a digital library of complex scientific multimedia is examined as an alternative to the Fedora Content Model (FCM). Using METS as an “intermediary” schema, where it functions as a template that is populated with content metadata on the fly using Extensible Stylesheet Language Transformations (XSLT), it is possible to replicate the flexibility of structure and granularity of FCM while avoiding its complexity and often substantial demands on developers. METS as an Intermediary Schema for a Digital Library of Complex Scientific Multimedia Of the many possible approaches to structuring complex data for delivery via the web, two divergent philosophies appear to predominate. One, exemplified by such standards as the Metadata Encoding and Transmission Standard (METS)1 or the Digital Item Declaration Language (DIDL),2 relies on the structured packaging of the multiple components of a complex object within “top-down” hierarchies. The second, of which the Fedora Content Model (FCM) is perhaps a prime example,3 takes the opposite approach of disaggregating structural units into atomistic objects, which can then be recombined according to the requirements of a given application.4 Neither is absolute in its approach—METS, for instance, allows cross-hierarchy linkages, and many FCM models are designed hierarchically—but the distinction is clear. Many advantages are validly claimed for the FCM approach to structuring digital data objects. Individual components, not constrained to hierarchies, may be readily reused in multiple representations with great flexibility.5 Complex interobject relationships may be encoded using semantic linkages,6 a potentially much richer approach to expressing these than the structural relationships of XML can allow. Multiple levels of granularity, from that of the collection as a whole down to its lowest-level components, can readily be modelled, allowing interobject relationships to be encoded as easily as intercomponent ones.7 Such models, particularly the RDF-based Fedora content model, are very powerful and flexible, but can often lead to complexity and consequently considerable demands on system development before they can be implemented. In addition, despite the theoretical interoperability offered by RDF, in practice the exchange and reuse of content models has proved somewhat limited because considerable work is usually required to re-create and validate a content model created elsewhere.8 This article examines whether it is possible to replicate the advantages of this approach to structuring data within the constraints of the more rigid METS standard. The data used for this analysis is a set of digital objects that result from biological nanoimaging experiments, the interrelationships of which present complex problems when they are delivered online. The Richard Gartner (richard.gartner@kcl.ac.uk) is a Lecturer in Library and Information Science, King’s College, London. METS AS AN INTERMEDIARY SCHEMA FOR A DIGITAL LIBRARY OF SCIENTIFIC MULTIMEDIA | GARTNER 25 method used is an unconventional use of a METS template as an intermediary schema;9 this allows something of the flexibility of the FCM approach while retaining the relative simplicity of the more structured METS model. A Nanoimaging Digital Library and its Metadata Requirements The collection analysed for this study derives from biological nanoimaging experiments undertaken at the Randall Division of Cell and Molecular Biophysics at King’s College London. Biological nanoimaging is a relatively new field of research that aims to unravel the biological processes at work when molecules interact in living cells; this is done by using optical techniques that can resolve images down to the molecular level. It has particular value in the study of how diseases progress and has great potential to help predict the effects of drugs on the physiology of human organs. As part of the Biophysical Repositories in the Lab (BRIL) project at King’s College London,10 a digital library is being produced to meet the needs of practitioners of live cell protein studies. Although the material being made available here is highly specialised, and the user base is restricted to a specialist cohort of biologists, the challenges of this library are similar to those of any collection of digital objects: in particular, the metadata strategy employed must be able to handle the delivery of complex, multifile objects as efficiently as, for example, a library of digitized books has to manage the multiplicity of image files that make up a single digital volume. The digital library itself is hosted on the widely used Fedora repository platform; as a result, it is employing FCM as the basis of its data modelling. The purpose of this analysis is to ascertain whether METS can also be used for the complex models required by this data and to compare its potential viability as an architecture for this type of application with FCM. A particular challenge of this collection is that the raw images from which it is constituted require combining and processing before they are delivered to the user. A further challenge is that the library encompasses images from a variety of experiments, all of which combine these files in different ways and employ different software for processing them. Some measure of the complexity of these requirements can be gathered from figure 1 below, which illustrates the processes involved in delivering the digital objects for two types of experiments. Figure 1. Architecture for Two Experiment Types METS AS AN INTERMEDIARY SCHEMA FOR A DIGITAL LIBRARY OF SCIENTIFIC MULTIMEDIA | GARTNER 26 The images created by two experiments, bleach and ACTIN_5, are shown here: it will be seen that the bleach experiment is divided into two subtypes (here called 2Grating and Apotone). Each type or subtype of experiment has its own requirements for combining the images it produces before they are displayed. For the subtype 2Grating, for instance, two images, each generated using a different camera grating, are processed in parallel (indicated by the brackets); these are then combined using the software package process-non-clem (shown by the hexagonal symbol) to produce a display image in TIFF format. The subtype Apotone requires three grating images and a further image with background information to be processed in parallel by the software process-apotone; in this case, the background image provides data to be subtracted from the combined three grating images to produce the final TIFF for display. ACTIN_5 experiments are entirely different: they produce still images that need to be processed sequentially (shown by the braces) to produce a video. Encoding the BRIL Architecture in METS This architecture, although complex, is readily handled within METS in a manner analogous to that of more conventional collections. As in any METS file, the structure of the experiments, including their subexperiments, is encoded using nested division (