Unveiling 50 years of librarianship research: An innovative study using unsupervised AI techniques

Abstract

In this study, 3646 articles published in the Scopus database between 1975 and 2024 were analyzed to examine the main research themes, trends, and research gaps in the field of librarianship. Using Latent Dirichlet Allocation (LDA) implemented in Python (Gensim, spaCy, pyLDAvis), a 17-topic model with a coherence score of 0.5266 was generated. The preprocessing steps included tokenization, lemmatization, bigram/trigram construction, and the removal of both general and domain-specific stopwords. The bibliometric analysis showed that most publications were classified under Social Sciences (93%), with the USA (1961 articles) being the leading contributor. Among the identified themes, Information, Research, and Technology (10.59%), Students' Information Literacy (9.22%), and Information Development (8.23%) emerged as the most prominent, while Professional Identity and Practice (3.37%) and Education and Research Competency (3.65%) were underrepresented. Trend analyses revealed that technology-related topics have accelerated significantly since 2010, reflecting the digital transformation of librarianship. In contrast, traditional themes such as Information Development and Health and Medical Services have declined in momentum. The study shows that digitalization and information literacy remain dominant in the field. In addition, it concludes that significant research gaps exist in professional identity, education, and competency development.