Introduction

Libraries have been exploring new ways to make their content available in digital formats for decades. They host all types of rich content, including images, text, newspapers, postcards, maps, journals, video and audio. Reuse is defined as use of the data regardless of when it is used, the purpose, the characteristics of the data and its user (van de Sandt et al., 2019). A common case of reuse is when an institution publishes a dataset (e.g. a bibliographic catalogue) and a researcher reuses this data for a particular purpose (e.g. assessing its quality). New initiatives have emerged to publish and reuse digital content in recent years. For example, the International GLAM Labs Community (https://glamlabs.io) promotes the creation of new teams in cultural heritage institutions that focus on the publication, under open licences, and reuse of digital collections in innovative and creative ways (Mahey et al., 2019). In a similar way, the Collections as Data initiative aims to promote the publishing of digital collections suitable for responsible computational use (Padilla et al., 2023). In the GLAM context, compu - tational use refers to the use of advanced techniques such as computer vision, named-entity recognition, information retrieval or network analysis. Several tools, including checklists and workflows, were recently published in order to facilitate the adoption of collections as data approaches in different contexts, such as small and medium-sized institutions (Candela et al., 2023), and more advanced environments, such as data spaces (Candela, Chambers and Irollo, 2023). Other initiatives are focused on the publication of FAIR (findable, accessible, interoperable and reusable) data (Wilkinson et al., 2016) that are complemented by the CARE principles to encourage ethical data use (Carroll et al., 2020).

Introduction

In this chapter, we discuss the application of corpus linguistic approaches to printed catalogue data and how they open up new areas of enquiry into cataloguing and curatorial practice over time. While similar methods are being adopted by scholars and practitioners in a range of fields, here we are primarily concerned with research in libraries that aims to develop collections metadata processes and practice for description, discovery, enrichment and computational use.

Our use case is Atanassova's project ‘Legacies of Curatorial Voice in the Descriptions of Incunabula Collections at the British Library and Their Future Reuse’, undertaken between 2022 and 2023, with Baker as her mentor. This research was funded under the Research Libraries UK and Arts and Humanities Research Council Professional Practice Fellowship scheme that enables library professionals to be active participants in research, and it was one of ten funded projects in the first year of the scheme (Research Libraries UK, 2024). The project took a ‘catalogues as data’ approach by using corpus linguistics methods to analyse legacy catalogue descriptions of books printed before 1500, with the ambition of gaining new insights into both the catalogues and the collections they describe. The motivation for undertaking this project was to use and embed the research methodology within British Library (BL) operations as a means of demonstrating the impact of digital scholarship applied to collection catalogues.

Introduction

In recent years, as digital textual analysis methods have become more userfriendly, and as access to the computing power necessary to run these analyses has become more readily available, there has been a surge of interest in creating textual corpora – collections of digitised, machine-processable texts – and doing what is sometimes referred to as ‘distant reading’ on them: analysing these corpora quantitatively in search of patterns (Underwood, 2017). In some examples of this sort of research, the corpus is well-defined and clearly bounded, (e.g. all published works by a particular author). In other cases, the composition of the corpus is more subjective; for example, an analysis of 240 crime novels in which one of four specified cities ‘play[s] a major role’ (Rauscher et al., 2013, 65). More problematic examples are those where the composition of the underlying corpus is unknown, particularly the hundreds of published papers based on the Google Books Ngrams data (books.google.com/ngrams), which traces the frequency of word usage over time among the vast and largely undocumented collection of books that make up the Google Books database. As has been demonstrated (e.g. Pechenick, Danforth and Dodds, 2015), the Google Books corpus is skewed towards academic publications rather than popular works, which makes it dangerous to draw conclusions from it about changes in popular culture or popular language usage.

Introduction

Scholarship on library catalogues and cataloguing practices provides numerous examples of how catalogue metadata reflect social biases regarding gender, racialised ethnicity, religion and sexuality, among others. This chapter takes a higher level view of library catalogues’ metadata, examining the numerous actors and practices that influence the creation and interpretation of catalogue metadata. Drawing on the work of critical heritage studies and feminism scholars in Europe, North America and Australia, I discuss the power exerted in the creation of catalogue metadata and the implications of such power relationships for research using these metadata. Acknowledging the inevitability of bias in data, I describe opportunities to manage catalogue metadata biases using natural language processing (NLP). Though bias cannot be eliminated, when critically applied, NLP methods can make the perspectives included and excluded in a catalogue explicit, informing research that aims to understand, use and improve library catalogues’ metadata and, in turn, the information discovery process. Due to overlaps in the cataloguing practices of GLAM institutions (galleries, libraries, archives and museums), this chapter refers to work relevant to the GLAM sector as a whole in addition to libraries specifically.

Digital and online technologies are enabling broader access to heritage in long-established libraries, such as the National Library of Scotland, through its Data Foundry platform (https://data.nls.uk).

Introduction

Managing extensive data has become an increasingly important part of information work since digital modes of storage and manipulation became widely available during the last decades of the 20th century. The sense that we have transitioned into a data-driven digital age sometimes leads to a present-minded view that positions library data as a contemporary phenomenon, of interest principally for tracking the needs and interests of current users and professionals. However, libraries have always relied on structured data to manage their collections. While this data has proliferated and become more tightly networked in modern contexts, it has done so based on practices with deep historical roots. Records ranging from the fragmentary and casual to the punctilious and voluminous have been employed for logging orders, tracing purchases, organising shelving, allowing users to locate materials and tracking responsibility for books, both within collections and when volumes depart from library premises with borrowers. The modules of modern library software, such as Ex Libris Group's Voyager, covering processes including acquisitions, cataloguing, circulation, reservations and various forms of reporting, are digital instantiations of what, in older libraries, were paper lists, bound accounts, laboriously-compiled manuscript catalogues, hefty receipt books and often snippy records of collection audits (it is very common for 18th-century curatorial documents from Scottish university libraries to bemoan the poor condition of books and uneven compliance with requests for returns from the more senior members of the university community).

Introduction

This chapter illustrates the feasibility of analysing large-scale intellectual and cultural patterns through bibliographic data. Our focus is to adopt a statistical lens to examine printing practices and to start thinking about how this correlates with reading habits in early modern Europe. We analyse the cultural impact of book formats, provide a quantitative evaluation of these formats and briefly explore the implications of these shifts on printing and reading practices during the early modern period. Additionally, we consider how 18th-century authors, like David Hume, responded to these general trends. The case studies of printing Shakespeare and the Bible will anchor our analysis. While extensive research has been dedicated to the intellectual and cultural significance of Shakespeare and the Bible, there have been relatively few efforts to statistically evaluate their publication trends, including changes in book formats and overall book size over time.

The late 18th century entailed a change in the reading and writing of books. Rolf Engelsing famously suggested a reading revolution took place in the second half of the century in which the public gradually switched from reading a few key works intensively and repeatedly, such as the Bible, to reading extensive amounts of literature by different authors (Engelsing 1970; 1974, 1–4). The breakthrough of extensive reading and the ensuing exposure to a broader spectrum of ideas is one of the key factors driving modernity. However, assessing these links requires better knowledge of when and where reading habits changed.

As the editors of this volume have attested, library catalogues have existed in electronic form since at least the 1960s. The transition from analogue to electronic catalogue significantly strengthened the ability of libraries the world over to administer collection development and access at the local, regional, national and even international level. While the electronic catalogue has long helped libraries realise a range of administrative functions, it has been less common for the library catalogue to be treated as data for research inherently suited to disciplinary inquiry that leverages computational methods at macro and microscopic scales. A library catalogue is not an archive, but it bears a similarity in the sense that it records the organic activities of an organisation that just happens to shape epistemic and ontological awareness through its regular operations. A library catalogue is an accretion, a fossil record of subjective decisions that evidence what an organisation operating in a particular sociocultural context believes is important and, consequently, what is not important. The library catalogue channels contested collection descriptions that directly influence how communities encounter and assess knowledge, circulation data complicates any sense of certainty with respect to the zeitgeist, and in the aggregate, library catalogues provide a dynamic map that characterises the waxing and waning of knowledge production globally. Library Catalogues as Data: Research, Practice and Usage introduces a diverse range of contributions that speak directly to these affordances and more.

Introduction

Large language models: Transforming artificial intelligence Since the release of BERT (Bidirectional Encoder Representations from Transformers) by Google in 2018 and GPT-3 by OpenAI in 2020, large language models (LLMs) have revolutionised the field of artificial intelligence (AI) and natural language processing. BERT pioneered the use of bidirectional transformers, which understand words by considering context from both sides, enabling a better grasp of natural language (Devlin et al., 2019). GPT-3 advanced this technology further with 175 billion parameters, providing unprecedented capabilities in text generation and comprehension (Frantar et al., 2023). These models are trained on vast amounts of textual data using deep learning techniques that capture the complex patterns and structures of human language. In analysing text sequences, LLMs use the transformer architecture for deep learning networks, allowing models to attend to different parts of the text simultaneously, thus improving the coherence and relevance of the generated responses. This capability enhances generative AI by helping it to create contextually appropriate content.

The transformer architecture was introduced in the seminal paper ‘Attention is All You Need’ (Vaswani et al., 2017). It demonstrated how self-attention mechanisms could outperform traditional recurrent neural networks in natural language processing tasks. This significantly improved the accuracy and efficiency of language models and laid the groundwork for subsequent developments.