Development of a Recommendation System for Scientific Articles Based on Machine-learning Models
DOI:
https://doi.org/10.24160/1993-6982-2025-4-138-145Keywords:
scientific recommendation systems, content-based filtering, RoBERTa, KeyBERT, machine learningAbstract
The article describes a scientific recommendation system (SRecS) developed to achieve a more efficient search for relevant publications considering a rapid growth of scientific literature in the field of computer sciences. By employing the proposed SRecS, the problem of information overburden experienced by subject-matter experts is solved through implementing a content-oriented approach and applying personalized search technologies, in particular, through construction of a user profile.
In developing SRecS, modern machine learning techniques were utilized, including the use of the RoBERTa pre-trained language model for topical classification and the KeyBERT model for extracting key words from publications. By substituting the full text of a scientific article with its vector representation through the use of specially extracted keywords, the system needs less computational efforts for its operation and at the same time becomes more versatile in nature.
The SRecS architecture consists of several modules, including those for linking with open-access electronic libraries, for classification of articles, for generating keywords, for constructing the user profile, and a recommendation algorithm. Owing to such modular structure, the system features high flexibility and offers the possibility to increase its functionality without the need to significantly changing the system’s general structure.
The SRecS preliminary testing results have demonstrated high quality of recommendations produced for small research teams engaged in studies in the field of computer sciences. The system capabilities can be further extended by involving additional sources of scientific data, incorporating advanced deep learning methods for processing multimodal information, and implementing interpretability mechanisms to ensure that the recommendations produced will be both transparent and well-justified.
References
2. Shahid A. e. a. Insights into Relevant Knowledge Extraction Techniques: a Comprehensive Review // J. Supercomputing. 2020. V. 76. Pp. 1695—1733.
3. Google Scholar [Электрон. ресурс] https://scholar.google.com (дата обращения 21.03.2025).
4. Semantic Scholar [Электрон. ресурс] https://www.semanticscholar.org (дата обращения 21.03.2025).
5. Bert D.J. Pre-training of Deep Bidirectional Transformers for Language Understanding [Электрон. ресурс] https://arxiv.org/pdf/1810.04805 (дата обращения 21.03.2025).
6. Liu Y. e. a. RoBERTa: a Robustly Optimized BERT Pretraining Approach [Электрон. ресурс] https://arxiv.org/pdf/1907.11692 (дата обращения 21.03.2025).
7. Grootendorst M. MaartenGr/KeyBERT: BibTeX» [Электрон. ресурс] https://www.maartengrootendorst.com/blog/keybert/ (дата обращения 21.03.2025).
8. Oh H., Wonkeun J., Dongil K. Attention-based Sequential Recommendation System Using Multimodal Data [Электрон. ресурс] https://arxiv.org/pdf/2405.17959 (дата обращения 21.03.2025).
9. Ng Yiu-Kai. Research Paper Recommendation Based on Content Similarity, Peer Reviews, Authority, and Popularity // Proc. IEEE XXXII Intern. Conf. Tools with Artifical Intelligence. 2020. Pp. 47—52.
10. Manning C.D., Raghavan P., Schütze H. An Introduction to Information Retrieval Cambridge: Cambridge University Press, 2009.
11. Li Z., Zou X. A Review on Personalized Academic Paper Recommendation. Computer and Information Sci. 2019. V. 12. Pp. 33—43.
12. Minaee Sh. e. a. Deep Learning Based Text Classification: a Comprehensive Review Электрон. ресурс] https://arxiv.org/pdf/2004.03705 (дата обращения 21.03.2025).
13. ArXiv [Электрон. ресурс] https://arxiv.org (дата обращения 21.03.2025).
14. The ACM Digital Library [Электрон. ресурс] https://dl.acm.org (дата обращения 21.03.2025).
15. Большакова Е.И. и др. Автоматическая обработка текстов на естественном языке и анализ данных. М.: Высшая школа экономики, 2017.
16. Bai X. e. a. Scientific Paper Recommendation: a Survey // IEEE Access. 2019. V. 7. Pp. 9324—9339.
17. Li Z., Zou X. A Review on Personalized Academic Paper Recommendation // Computer and Information Sci. 2019. V. 12. Pp. 33—43.
18. Kreutz C.K., Schenkel R. Scientific Paper Recommendation Systems: a Literature Review of recent Publications [Электрон. ресурс] https://www.researchgate.net/publication/357552714_Scientific_Paper_Recommendation_Systems_a_Literature_Review_of_recent_Publications (дата обращения 21.03.2025).
19. Gündogan E., Kaya M. A Novel Hybrid Paper Recommendation System Using Deep Learning // Scientometrics. 2022. V. 127(7). Pp. 3837—3855.
20. Козлов П.А. и др. Сравнительный анализ бинарных классификаторов на массиве научных публикаций // Заводская лаборатория. Диагностика материалов. 2022. V. 88(7). Pp. 79—87.
---
Для цитирования: Назаров Н.А. Разработка рекомендательной системы научных статей на основе моделей машинного обучения // Вестник МЭИ. 2025. № 4. С. 138—145. DOI: 10.24160/1993-6982-2025-4-138-145.
#
1. Bornmann L., Mutz R. Growth Rates of Modern Science: a Bibliometric Analysis Based on the Number of Publications and Cited References. 2015;66(11):2215—2222.
2. Shahid A. e. a. Insights into Relevant Knowledge Extraction Techniques: a Comprehensive Review. J. Supercomputing. 2020;76:1695—1733.
3. Google Scholar [Elektron. Resurs] https://scholar.google.com (Data Obrashcheniya 21.03.2025).
4. Semantic Scholar [Elektron. Resurs] https://www.semanticscholar.org (Data Obrashcheniya 21.03.2025).
5. Bert D.J. Pre-training of Deep Bidirectional Transformers for Language Understanding [Elektron. Resurs] https://arxiv.org/pdf/1810.04805 (Data Obrashcheniya 21.03.2025).
6. Liu Y. e. a. RoBERTa: a Robustly Optimized BERT Pretraining Approach [Elektron. Resurs] https://arxiv.org/pdf/1907.11692 (Data Obrashcheniya 21.03.2025).
7. Grootendorst M. MaartenGr/KeyBERT: BibTeX» [Elektron. Resurs] https://www.maartengrootendorst.com/blog/keybert/ (Data Obrashcheniya 21.03.2025).
8. Oh H., Wonkeun J., Dongil K. Attention-based Sequential Recommendation System Using Multimodal Data [Elektron. Resurs] https://arxiv.org/pdf/2405.17959 (Data Obrashcheniya 21.03.2025).
9. Ng Yiu-Kai. Research Paper Recommendation Based on Content Similarity, Peer Reviews, Authority, and Popularity. Proc. IEEE XXXII Intern. Conf. Tools with Artifical Intelligence. 2020:47—52.
10. Manning C.D., Raghavan P., Schütze H. An Introduction to Information Retrieval Cambridge: Cambridge University Press, 2009.
11. Li Z., Zou X. A Review on Personalized Academic Paper Recommendation. Computer and Information Sci. 2019;12:33—43.
12. Minaee Sh. e. a. Deep Learning Based Text Classification: a Comprehensive Review Elektron. Resurs] https://arxiv.org/pdf/2004.03705 (Data Obrashcheniya 21.03.2025).
13. ArXiv [Elektron. Resurs] https://arxiv.org (Data Obrashcheniya 21.03.2025).
14. The ACM Digital Library [Elektron. Resurs] https://dl.acm.org (Data Obrashcheniya 21.03.2025).
15. Bol'shakova E.I. i dr. Avtomaticheskaya Obrabotka Tekstov na Estestvennom Yazyke i Analiz Dannykh. M.: Vysshaya Shkola Ekonomiki, 2017. (in Russian).
16. Bai X. e. a. Scientific Paper Recommendation: a Survey. IEEE Access. 2019;7:9324—9339.
17. Li Z., Zou X. A Review on Personalized Academic Paper Recommendation. Computer and Information Sci. 2019;12:33—43.
18. Kreutz C.K., Schenkel R. Scientific Paper Recommendation Systems: a Literature Review of recent Publications [Elektron. Resurs] https://www.researchgate.net/publication/357552714_Scientific_Paper_Recommendation_Systems_a_Literature_Review_of_recent_Publications (Data Obrashcheniya 21.03.2025).
19. Gündogan E., Kaya M. A Novel Hybrid Paper Recommendation System Using Deep Learning. Scientometrics. 2022;127(7):3837—3855.
20. Kozlov P.A. i dr. Sravnitel'nyy Analiz Binarnykh Klassifikatorov na Massive Nauchnykh Publikatsiy. Zavodskaya Laboratoriya. Diagnostika Materialov. 2022;88(7):79—87. (in Russian)
---
For citation: Nazarov N.A. Development of a Recommendation System for Scientific Articles Based on Machine-learning Models. Bulletin of MPEI. 2025;4:138—145. (in Russian). DOI: 10.24160/1993-6982-2025-4-138-145
