Topic Modelling in Library and Information Science from the Primary Data: Swing in Thrust Areas

This paper portrays the topic modeling approach in identifying critical facets in new Library and Information Science (LIS) research. Using machine language, topic modeling in statistics recognizes the "topics" in a document pool. Latent Dirichlet Allocation (LDA) is one such model for topic analysis and thus is used to classify text in the literary warrant in LIS in various locations worldwide. The study analyzed 734 articles published in 4 different LIS journals, i.e., Annals of Library and Information Studies (Indian), DESIDOC Journal of Library and Information Technology (DJLIT-Indian), College and Research Libraries (International), Liber Quarterly (International) from 2016 to 2020. LDA is used to analyze and manage the dataset. The study pivoted on the foundation of extracting 'topics' from these four journals using Mallet and Knime software to determine what research topics are essential at the national and global levels. The paper also visualizes a relationship map of distributed issues at the national and international levels to identify the sublime contexture of keyed-in tags. Furthermore, the study took a socio-cultural approach to LIS research with cross-country comparisons. The study found that the compared topics retrieved from both journals published in India and abroad where non-Indian journals varied significantly from those published in Indian journals. The study concluded that cross-cultural adoption is visible in the research activities over time. Moreover, this study identified the social parameters that describe the changes in the discipline.