Low-Cost AI/ML Framework for Classifying Indian Research Publications
DESIDOC Journal of Library & Information Technology · 13.11.2026 · 46 (6) · 743-754
DOI: 10.14429/djlit.22059 ↗Kurz gefasst
Fachartikel in der Zeitschrift DESIDOC Journal of Library & Information Technology, online erschienen am 13.11.2026. Verfasst von Anjan Pal und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 2.5).
Abstract (Originalsprache)
Analysing the research performance of a country or institution by the SDG alignment of its publications is increasingly important, as the Times Higher Education sustainability impact ranking since 2019 shows. Manual classification of growing scholarly output by Sustainable Development Goals is nearly impossible and commercial automated systems can be costly. This study proposes a low-cost AI/ML framework for automated SDG classification of Indian research publications using open-source tools (Annif, OpenRefine) and open datasets (OpenAlex, Aurora). We retrieved 765,672 Indian publications (2016-2025) from OpenAlex, curated 752,355 records and built a balanced dataset of 248,992 records. Five Annif backends were evaluated using F1@k and NDCG. Statistical validation (Friedman, Wilcoxon-Holm, Cohen’s d) established a performance hierarchy. A Neural Network ensemble of SVC, FastText and OmikujiB significantly outperformed individual models (p1.3). On the human-validated OSDG Community Dataset (1,536 records), our CPU-based prototype, trained on only 200K records, achieved 62.2 % top-1 accuracy, against 62.5 % for Aurora, a GPU-based system trained on millions of documents. The framework makes SDG monitoring accessible to resource-limited Indian institutions.
Zitieren
APA
Pal, A., & Mukhopadhyay, P. (2026). Low-Cost AI/ML Framework for Classifying Indian Research Publications. DESIDOC Journal of Library & Information Technology, 46(6), 743-754. https://doi.org/10.14429/djlit.22059
Vancouver
Pal A, Mukhopadhyay P. Low-Cost AI/ML Framework for Classifying Indian Research Publications. DESIDOC Journal of Library & Information Technology. 2026;46(6):743-754. doi:10.14429/djlit.22059
BibTeX
@article{Pal2026,
author = {Pal, Anjan and Mukhopadhyay, Parthasarathi},
title = {Low-Cost AI/ML Framework for Classifying Indian Research Publications},
journal = {DESIDOC Journal of Library & Information Technology},
year = {2026},
volume = {46},
number = {6},
pages = {743--754},
doi = {10.14429/djlit.22059}
}
EUVIMED