Navigating the artificial intelligence frontier on cataloguing and metadata work in libraries: an interview with Getaneh Alemu
Year: 2025
Type: article
Abstract: Getaneh Alemu is an experienced Cataloguing and Metadata Librarian with over two decades of experience in higher education and libraries. He specialises in ensuring that metadata is accurate, consistent, usable and standardised. His expertise covers several key metadata standards and tools, including MARC21, MarcEdit, Dublin Core, AACR2, RDA, LCSH, DDC, Alma and Primo. These standards aim to maximise the visibility, discoverability, accessibility and utilisation of library resources.Getaneh has an impressive academic background that includes a PhD in Metadata for Digital Libraries from the University of Portsmouth, earned in 2014. He also holds an MA in Digital Library Learning from Oslo University College, Parma University and Tallinn University, obtained in 2009. Additionally, he earned an MA in Knowledge and Information Management from Ghent University in 2005 and a BA in Library and Information Science from Addis Ababa University in 2001.His research on metadata is well-documented through key publications such as “Metadata Standards and Models” (2024), “The Future of Enriched, Linked, Open and Filtered Metadata” (2022) and “An Emergent Theory of Digital Library Metadata: Enrich then Filter” (2015).Artificial intelligence (AI) is poised to automate repetitive, data-intensive and pattern-recognition tasks, thereby freeing human cataloguers for more complex work. What impact will this have?We interviewed Getaneh Alemu because his current work demonstrates a proactive engagement with AI in cataloguing. He is actively involved in exploring, testing and evaluating various AI tools. This exploration extends to discussions with colleagues to understand AI’s practical applications in metadata creation, enrichment and the enhancement of library catalogues and discovery systems.Q1.How does Artificial Intelligence currently impact your work as a cataloguer?Currently, the impact of AI on my cataloguing work is mainly in the exploratory phase. I am actively exploring, testing and evaluating various tools, including the new AI Metadata Assistant from Ex Libris, as well as Copilot, ChatGPT and Gemini. Additionally, I am engaged in ongoing discussions with colleagues in the library to gain a better understanding of the practical applications of these technologies in metadata creation and enrichment, as well as in improving the library catalogue and discovery.We are entering an exciting phase and we must engage as active participants rather than as complainers or passive observers. It’s important to remember that AI is not an external force intruding upon our lives; rather, it is a human innovation that we must use wisely, ethically and responsibly. AI is no longer a distant concept, relegated to the realm of science fiction; it is a reality. The emergence of large language models like ChatGPT and Copilot has made AI increasingly relevant to our work, transforming it from an abstract idea into a practical tool for organising and managing bibliographic data. This process is ongoing and requires continuous learning, critical reflection and experimentation, but is happening at a fast pace.I stay informed about the latest developments in Artificial Intelligence by reading authoritative sources, including “New Horizons in Artificial Intelligence in Libraries” (Balnaves et al., 2025), as well as peer-reviewed journal articles and professional discussions on platforms like LinkedIn. In my workplace, a dedicated team of information librarians regularly updates us on new AI tools and innovations. In this regard, my university shows significant proactivity and remains at the forefront of technological advancement. The university has provided a concise online training module covering the fundamental principles of artificial intelligence, which I have successfully completed.Currently, the impact of AI on my professional responsibilities is limited but overall positive. Although its use is not yet widespread, I am involved in a small-scale experimental project that explores the use of AI-generated metadata for subject headings. We plan to provide further updates as this project progresses.Q2. What are some examples of successful AI implementations in metadata management and cataloguing?We’re observing some promising developments, although they are still in the experimental stages. I utilise the Copilot and Gemini AI tools to check Dewey Decimal numbers and to suggest appropriate subject headings. Ex Libris has introduced an AI assistant called “Metadata Enrichment using AI,” which aids in the ongoing enhancement of bibliographic records. It’s essential to note that Ex Libris emphasises the need for human cataloguers to validate the outputs generated by AI to ensure accuracy and trustworthiness.I tested the AI Assistant with both new and existing records to evaluate its capability to create MARC 21 bibliographic records from minimal input. The tool can handle basic bibliographic elements (such as title and author) or visual sources (like a book cover or an image of the table of contents) and generates a draft record that is sent to the Metadata Editor (MDE).For this test, I supplied an image sourced from a publisher’s website, providing the main title, author and year but deliberately omitting the subtitle. The AI inferred and correctly populated the title along with the subtitle, inserted the correct MARC21 245 field, auto-assigned some basic (generic) subject headings (in the 650 LCSH field) and generated a summary note (520). It also embedded provenance metadata via a note indicating: “Part of the metadata in this record was created with the help of the AI Metadata Assistant.” However, some essential metadata elements were missing such as physical description (300), classification numbers (082 / 050), standard identifier and richer (granular) subject access. ISBD punctuation (e.g. colons, commas) and subfield spacing in 245 and elsewhere were also inconsistent. While effective for generating core descriptive metadata, the Assistant requires cataloguer input to ensure full bibliographic control, structural accuracy and standards compliance.AI has significant potential to improve metadata and cataloguing. While it offers powerful capabilities, integrating AI with existing standards and tools requires effort. AI can enhance metadata, providing users with more intuitive access to information through linked subject systems. Additionally, it can address historical biases in classification systems and improve data quality by conducting consistency checks and detecting anomalies in bibliographic records.Q3.What are your thoughts on the risks and ethical considerations of using AI in metadata and cataloguing?AI-augmented metadata must adhere to established ethical principles by prioritising user-centred access and non-discrimination, reflecting IFLA’s emphasis on “user convenience.” This requires actively addressing the inherent biases present in both AI algorithms and the socio-cultural biases that are built into historical metadata standards. As noted by Gartner (2016) and Bowker and Star (1999), metadata is never completely neutral; every standard and category reflects a particular viewpoint while silencing others. This can result in historical classification schemes that represent colonial perspectives or lack appropriate terminology for diverse cultural expressions.To combat these entrenched biases, it is essential to continuously and critically examine, diversify and decolonise metadata practices to ensure inclusivity and equitable representation.Thus, transparency and accountability in the metadata decisions made by AI are crucial. It is important to highlight the vital role of human oversight and the need for collaborative efforts among all metadata professionals to ensure ethical and responsible information practices.Q4.To what extent can AI automate cataloguing tasks and what tasks are best suited for automation?AI holds significant potential to automate various cataloguing tasks, thereby increasing efficiency and freeing up human cataloguers for more complex, nuanced work. The tasks best suited for automation generally involve repetitive, data-intensive, or pattern-recognition processes.One key area is data remediation and normalisation. AI can effectively identify and rectify common errors or inconsistencies present in legacy metadata records, particularly useful for older data, such as pre-RDA records created under AACR2, which often contain numerous abbreviations and display inconsistencies that can be automatically corrected. Similarly, AI excels at format conversion, efficiently transforming metadata from one standard to another, for instance, from MARC21 to BIBFRAME, XML, or compatible formats for archival metadata.AI can also significantly enhance record enrichment and cross-linking. It can automatically augment bibliographic records by adding granular details and establishing crucial connections. This involves incorporating URIs from authoritative sources such as Wikidata, LCNAF and VIAF, thereby ensuring accurate and unique identification of entities like authors, editors, agencies and events, which is vital for achieving full Linked Data compliance. Furthermore, given the exponential growth of e-book, database and e-journal collections, AI is invaluable for identifying defunct links by scanning e-resource records to flag broken or outdated URLs, proactively resolving access issues and ensuring consistent full-text access for users.Other highly suitable tasks for automation include bulk updates and imports, allowing for large-scale changes on bibliographic records or efficient ingestion of new records. AI can also assist in comparing bibliographic record sets, automatically finding common or differing records, similar to an Excel VLOOKUP function. For the initial stages of cataloguing, AI can be instrumental in automatic record generation, creating preliminary bibliographic records directly from full-text books and documents, which human cataloguers can then review.Beyond structured data, AI’s capabilities extend to deciphering illegible texts, extracting information from old handwritten archival materials. It can also assist with intellectual analysis by suggesting relevant subject headings based on content analysis. Finally, AI can ensure records adhere to specific presentation standards like ISBD through automated formatting and indentation and provide initial translations of metadata from one language to another, requiring subsequent human review for accuracy and nuance.Q5. How can AI enhance the discoverability and accessibility of library resources?When integrated with existing standards, systems and practices, AI enhances bibliographic records, improving discoverability and accessibility. It strengthens the core functions of metadata as defined by standards like FRBR, IFLA LRM and RDA. These standards are designed to help users find information, identify relevant resources, select appropriate materials, gain access and explore related content (Riva, Le Boeuf and Žumer, 2017). By harnessing enriched, linked, enhanced and open metadata, AI amplifies these capabilities. It achieves this through automating processes such as content indexing, metadata creation, labelling and summarisation, which makes resources more accessible across various disciplines and contexts. AI also improves information retrieval and analysis, allowing users to find and analyse relevant information within large and complex data sources more effectively. This facilitates better access to extensive collections of data through advanced techniques like intelligent grouping and filtering.Ultimately, AI enables large-scale global access through automated classification, organisation and retrieval. It also offers multi-dimensional subject access through innovations like topic modelling and AI assistants, fostering a richer, interconnected and more equitable learning environment for users.Q6.How can libraries ensure data quality and accuracy when using AI for cataloguing?Ensuring data quality and accuracy is crucial. AI models rely on exceptionally large data sources to learn patterns, highlighting the importance of providing high-quality input data. Discussions about algorithmic bias, as well as the need for transparency and explainability in machine learning, are highly relevant; AI requires accurate provenance metadata. Consequently, librarians must verify metadata for accuracy and consistency. To achieve this, specific targets should be set, quality metrics established and human oversight implemented to maintain the accuracy and consistency of AI-generated metadata.Q7. Can technology replace the cataloguer? Why or why not?My answer is no, and here is why. Technology and innovation are not threats to librarianship but rather indispensable allies in its ongoing evolution. As a profession, we stand on the shoulders of visionary librarians such as Panizzi, Ranganathan, Dewey, Cutter, Garfield, Avram, Otlet and many others, each of whom embraced innovation and the tools of their time to reimagine new ways of thinking, working and information organisation. Their work exemplifies the dynamic and adaptive spirit at the heart of librarianship.Today, with the advent of AI, we are charting a new horizon, one marked not by displacement but by transformation. I am unequivocally in favour of AI, not as a replacement for human expertise but as a catalyst for efficiency, enrichment and improved discovery. We, like our predecessors, must embrace adaptation, readily adjusting our approaches with ethical discernment and unwavering professional integrity. In doing so, we perpetuate librarianship’s legacy as a field that is both profoundly humanistic and dynamically responsive to technological advancements.The future of metadata creation is evolving, blending human expertise with the computational efficiency of AI, as discussed in “New Horizons in Artificial Intelligence in Libraries” (Balnaves et al., 2025). This integration fosters richer and more inclusive metadata ecosystems. In this model, metadata creators, users and AI will collaborate closely. The hybrid approach of co-creating metadata is essential, as human judgement is crucial for maintaining trust and accuracy in light of the limitations of AI, such as bias and factual errors. Instead of viewing technology as a replacement for cataloguers, we should embrace a future where human insight and machine capabilities work together to enhance bibliographic metadata.Q8.What skills will be essential for cataloguers in an AI-driven environment?We must continually explore, learn and better understand AI, actively engaging with it rather than remaining on the sidelines. Being proactive and hands-on involves mastering prompt engineering, delving into the logic behind algorithms and fostering a richer, more engaging research and learning environment. Therefore, essential skills for cataloguers in an AI-driven environment will undoubtedly include comprehensive AI literacy, encompassing prompt engineering and a foundational grasp of AI principles. Deep subject analysis and cataloguing skills will remain paramount, enabling cataloguers to strategically cede repetitive tasks to AI. Furthermore, a strong knowledge of the ethical considerations surrounding AI, coupled with developed technical skills in digital and algorithmic literacy and competencies in developing and applying various AI approaches, will be crucial.Q9.What training or resources should be provided to cataloguers to prepare them for AI integration?To effectively prepare librarians and cataloguers for AI integration, training, development and resources are essential. This includes providing opportunities for direct experimentation with AI tools, such as chatbots, to build hands-on understanding of their capabilities and limitations. Access to open-source toolkits with guides and examples for recommender systems should be made available for those keen to explore coding applications. Furthermore, robust educational resources, like online courses covering AI basics and ethical implications, alongside current research reports on AI’s impact on information professions, are vital. Finally, encouraging and supporting proof-of-concept projects will allow cataloguers to develop crucial technical and business skills in a practical, real-world context. I believe we can all come together to teach one another, continuously experiment and improve our processes. We can achieve this through seminars, conferences, magazines, articles and books.Q10.What role does human oversight and intervention play in AI-assisted cataloguing?Human oversight and intervention are crucial, especially in AI-assisted cataloguing. This situation is reminiscent of the outdated belief that paperless offices would render libraries obsolete. Such hype often overlooks the importance of traditional institutions. Just as physical libraries, archives and museums remain vital despite decades of technological advancements, human involvement will continue to be central in the age of AI.In this era, cataloguers play an essential role in subject analysis and metadata creation. They interpret the content, context and purpose of resources, selecting appropriate subject headings while upholding the principles of Equity, Diversity and Inclusion (EDI). While AI can assist with tasks like keyword extraction, the nuanced process of subject analysis demands human judgement, cultural sensitivity and an understanding of complex standards such as Resource Description and Access (RDA) and IFLA’s Library Reference Model (LRM).Cataloguers are adept at distinguishing between different works and editions, often requiring professional judgement and background knowledge. Although AI can enhance efficiency, it lacks the critical thinking and ethical awareness that skilled cataloguers possess.Q11.How might AI change the future of metadata standards and practices?Efforts are being made to transition from outdated data formats such as MARC and AACR2 to modern standards like BIBFRAME and RDA. Similarly, we must envision and adapt our metadata standards and practices to keep pace with advancements in AI.AI’s influence on metadata is significant, as it will both utilise existing metadata and drive the creation of new standards. Metadata will be essential for establishing trust and authenticity in AI-generated content, ensuring provenance and integrity.Our standards will facilitate interoperability, fostering Linked Data environments, while also guiding ethical AI practices by incorporating usage rights and privacy considerations. As cataloguers, we must strategically integrate AI, recognising its potential for efficiency alongside the necessity for human oversight to address current imperfections like inaccurate data. This moment reflects a “paradigm shift” in how we approach metadata in light of AI’s capabilities.
Source: Digital Library Perspectives
Authors: Getaneh Alemu, Anna Maria Tammaro
Institutions: Southampton Solent University, University of Parma
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4413313520