Mapping knowledge patterns in IGNOU research using text mining and topic modeling

This study presents a comprehensive scientometric and text-mining analysis of the research output of Indira Gandhi National Open University (IGNOU) indexed in the Scopus database up to 2025, based on 2,254 publications. Bibliometric indicators were integrated with Latent Dirichlet Allocation (LDA) topic modeling, applied to titles and abstracts following a rigorous preprocessing pipeline including boilerplate removal, domain-specific stop-word filtering, lemmatisation, and bigram/trigram detection. Topic solutions from k = 2 to 15 were compared using C_v-proxy (NPMI) coherence, UMass coherence, and perplexity; a six-topic model was selected (C_v-proxy = 0.062; perplexity = 1,612.7) on coherence and interpretability grounds, then validated through ten-seed stability testing (mean top-word Jaccard ≈ 0.42, indicating moderate sensitivity to initialisation) and a four-configuration preprocessing sensitivity analysis. The six themes identified are Engineering, Renewable Energy, and Computational Systems; Biomedical, Pharmacological, and Molecular Sciences; Environmental, Geospatial, and Land-Use Studies; Agricultural and Plant Sciences for Sustainability; Management, Information Systems, and Industry Analytics; and Open and Distance Education Development. Temporal analysis shows that Open and Distance Education Development remains the largest and most consistent theme throughout 2000–2025, while the Engineering, Biomedical, and Agricultural themes grow fastest from 2020 onward. Citation impact proved theme-dependent rather than volume-dependent: Biomedical, Pharmacological, and Molecular Sciences recorded the highest average citation rate (12.79 per paper) despite modest volume, while Open and Distance Education Development, the largest theme, had the lowest (4.75 per paper). Open Access publications received 13.09 citations per paper on average versus 8.59 for non-Open Access, an approximate 52% advantage. These findings demonstrate the value of rigorously validated topic modeling combined with scientometric indicators for institutional benchmarking and strategic research planning. Findings are drawn from Scopus-indexed records only and rely on raw, non-field-normalised citation counts; the Open Access advantage reflects an association rather than a causal estimate, and the topic model, though now validated through systematic k-selection and stability/sensitivity testing, remains a data-driven approximation of thematic structure rather than a definitive classification.