Title Comparison Between SVM and DistilBERT for Multi-label Text Classification of Scientific Papers Aligned with Sustainable Development Goals
Authors Morales-Hernández R.C. , BECERRA ALONSO, DAVID, Vivas E.R. , Gutiérrez J.
External publication No
Means Lecture Notes in Computer Science
Scope Conference Paper
Nature Científica
JCR Quartile 4
SJR Quartile 3
SJR Impact 0.32
Web https://www.scopus.com/inward/record.uri?eid=2-s2.0-85142811549&doi=10.1007%2f978-3-031-19496-2_5&partnerID=40&md5=11e14ba1d222c2cec947c36e0e3c458d
Publication date 01/10/2022
Scopus Id 2-s2.0-85142811549
DOI 10.1007/978-3-031-19496-2_5
Abstract The scientific articles identification with the 17 sustainable development goals of the UN 2030 Agenda is a valuable task for research and educational institutions. Finding an efficient and practical multi-label classification model using machine or deep learning remains relevant. This work refers to the performance comparison of a text classification model that combines Label Powerset (LP) and Support Vector Machine (SVM) against a transfer learning language model such as DistilBERT in 5 different imbalanced and balanced dataset scenarios of scientific papers. A proposed classification process was implemented with performance metrics, which have confirmed that the combination LP-SVM continues to be an option with remarkable results in multi-label text classification. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
Keywords Classification (of information); Deep learning; Learning systems; Planning; Sustainable development; Text processing; Distilbert; Educational institutions; Label powerset; Multi-label text classification; Research institutions; Scientific articles; Scientific papers; Support vectors machine; Sustainable development goal; Transfer learning; Support vector machines
Universidad Loyola members

Change your preferences Manage cookies