gpntbНаучные и технические библиотекиScientific and Technical Libraries1027-36892686-8601Russian National Public Library for Science and Technology10.33186/1027-3689-2022-4-126-136gpntb-927Research ArticleИСКУССТВЕННЫЙ ИНТЕЛЛЕКТ В БИБЛИОТЕЧНОЙ ДЕЯТЕЛЬНОСТИARTIFICAL INTELLECT IN LIBRARIESНовый подход к процессу автоматизации обучения на основе данных о поведении пользователей в цифровых библиотекахNew approach to computer-aided learning based on digital library user behaviorКрупаТ. В.KrupaT. V.

Крупа Татьяна Викторовна – кандидат психологических наук, президент ООО «ГлобалЛаб»

Москва

Tatiana V. Krupa – Cand. Sc. (Psychology), President, Globallab Global Student Laboratory

Moscow

t.krupa@globallab.orgООО «ГлобалЛаб»РоссияGloballab Global Student LaboratoryRussian Federation20222604202204126136Copyright © Крупа Т.В., 20222022Крупа Т.В.Krupa T.V.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://ntb.gpntb.ru/jour/article/view/927

Представлена математическая модель применения рекуррентной сети с внешней памятью. Она предназначена для предсказания оптимальной образовательной траектории пользователя в цифровых информационных средах, к которым могут быть отнесены цифровые библиотеки. Основная задача, решаемая с помо щью метода машинного обучения, основанного на применении нейронных сетей, – индивидуализация образовательных траекторий пользователя. Цель работы – моделирование различных аспектов деятельности обучающегося с использованием рекуррентных нейронных сетей для более точной индивидуализации образовательной траектории. В основе метода лежат две разновидности рекуррентных нейронных сетей: классическая с сигмоидальной функцией активации и сеть с долгой краткосрочной памятью LSTM (Long Short-Term Memory). Результаты проведённых экспериментов показали существенные преимущества применения рекуррентных нейронных сетей для предсказания шагов образовательной траектории по сравнению с аналогичными методами. Таким образом, разработанная модель имеет более высокую точность предсказания (на 15–20% выше относительно аналогов). Основная область её применения – предсказание оптимальной образовательной траектории пользователя в цифровой информационной среде, в частности – цифровой библиотеке.

The author introduces the mathematical model of recurrent neural network with external memory. It is intended for predicting efficient education trajectory in digital information environments, e. g. digital libraries. The goal of computer-aided learning based on neural networks is to personalize user trajectories. In the study, user behavior is modeled for the more precise personalization in various aspects using recurrent neural networks. The method is designed for two types of recurrent neural networks, i. e. the classic one with sigmoidal activation function and that with LSTM (Long Short-Term Memory). The experiments demonstrated serious advantages of recurrent neural networks over analogous methods in predicting education trajectory. Thus, the proposed model is the more efficient in predictive accuracy (by 15–20% higher than analogous methods). Its prime application area is prediction of optimum user education trajectory in the digital information environment, and digital library, in particul

рекуррентная нейронная сетьРНСRNNобразовательная траекториямоделирование образовательной траекториицифровая библиотекамоделирование пользователей цифровых библиотекrecurrent neural networkRNNeducation trajectorydigital librarydigital libraryuser modelingПрикладное исследование, результаты которого изложены в настоящей статье, осуществлено при государственной финансовой поддержке Министерства образования и науки Российской Федерации в рамках соглашения № 14.576.21.0100 от 26 сентября 2017 г. (уникальный идентификатор – RFMEFI57617X0100).The article comprises the findings of the study completed through the state funding by the Ministry of Science and Higher Education of the Russian Federation under the agreement No. 14.576.21.0100 of September 26, 2017 (unique number – RFMEFI57617X0100).ReferencesPiech C. et al. Deep knowledge tracing // Advances in Neural Information Processing Systems. Stanford, 2015. C. 505–513. URL: http://papers.nips.cc/paper/5654-deep-knowledge-tracing (accessed: 21.04.2021).Piech C. et al. Deep knowledge tracing // Advances in Neural Information Processing Systems. Stanford, 2015. C. 505–513. URL: http://papers.nips.cc/paper/5654-deep-knowledge-tracing (accessed: 21.04.2021).Piech C. et al. Autonomously generating hints by inferring problem solving policies // Proceedings of the Second (2015) ACM Conference on Learning@ Scale. ACM, 2015. С. 195–204.Piech C. et al. Autonomously generating hints by inferring problem solving policies // Proceedings of the Second (2015) ACM Conference on Learning@ Scale. ACM, 2015. С. 195–204.Piech C. et al. Modeling how students learn to program // Proceedings of the 43rd ACM technical symposium on Computer Science Education. ACM, 2012. С. 153–160.Piech C. et al. Modeling how students learn to program // Proceedings of the 43rd ACM technical symposium on Computer Science Education. ACM, 2012. С. 153–160.Tang S., Peterson J. C., Pardos Z. A. Modelling Student Behavior using Granular Large Scale Action Data from a MOOC // arXiv preprint arXiv:1608.04789. 2016. URL: https://arxiv.org/abs/1608.04789 (дата обращения: 21.04.2021).Tang S., Peterson J. C., Pardos Z. A. Modelling Student Behavior using Granular Large Scale Action Data from a MOOC // arXiv preprint arXiv:1608.04789. 2016. URL: https://arxiv.org/abs/1608.04789 (дата обращения: 21.04.2021).Sayed M. et al. E-Learning optimization using supervised artificial neural-network / M. Sayed // Journal of software engineering and applications. 2015. Т. 8. № 1. С. 26. URL: http://file.scirp.org/Html/4-9302022_53428.htm (дата обращения: 21.04.2021). doi: http://dx.doi.org/10.4236/jsea.2015.81004.Sayed M. et al. E-Learning optimization using supervised artificial neural-network / M. Sayed // Journal of software engineering and applications. 2015. Т. 8. № 1. С. 26. URL: http://file.scirp.org/Html/4-9302022_53428.htm (дата обращения: 21.04.2021). doi: http://dx.doi.org/10.4236/jsea.2015.81004.Reddy S., Labutov I., Joachims T. Latent skill embedding for personalized lesson sequence recommendation // arXiv preprint arXiv:1602.07029. 2016. URL: https://arxiv.org/abs/1602.07029 (дата обращения: 01.05.2021).Reddy S., Labutov I., Joachims T. Latent skill embedding for personalized lesson sequence recommendation // arXiv preprint arXiv:1602.07029. 2016. URL: https://arxiv.org/abs/1602.07029 (дата обращения: 01.05.2021).Nerrand O. et al. Neural network training schemes for non-linear adaptive filtering and modelling // International Joint Conference on Neural Networks. 1991. Т. 1. С. 61–66.Nerrand O. et al. Neural network training schemes for non-linear adaptive filtering and modelling // International Joint Conference on Neural Networks. 1991. Т. 1. С. 61–66.Schmidhuber J. Deep learning in neural networks: An overview / J. Schmidhuber // Neural networks. 2015. Т. 61. С. 85–117. doi: 10.1016/j.neunet.2014.09.003.Schmidhuber J. Deep learning in neural networks: An overview / J. Schmidhuber // Neural networks. 2015. Т. 61. С. 85–117. doi: 10.1016/j.neunet.2014.09.003.Cader A. The Potential for the Use of Deep Neural Networks in e-Learning Student Evaluation with New Data Augmentation Method // International Conference on Artificial Intelligence in Education. Springer, Cham, 2020. С. 37–42.Cader A. The Potential for the Use of Deep Neural Networks in e-Learning Student Evaluation with New Data Augmentation Method // International Conference on Artificial Intelligence in Education. Springer, Cham, 2020. С. 37–42.Liu D. et al. Multiple Features Fusion Attention Mechanism Enhanced Deep Knowledge Tracing for Student Performance Prediction / D. Liu // IEEE Access. 2020. Т. 8. С. 194894–194903. doi: 10.1109/access.2020.3033200.Liu D. et al. Multiple Features Fusion Attention Mechanism Enhanced Deep Knowledge Tracing for Student Performance Prediction / D. Liu // IEEE Access. 2020. Т. 8. С. 194894–194903. doi: 10.1109/access.2020.3033200.Gervet T. et al. When is Deep Learning the Best Approach to Knowledge Tracing? / T. Gervet // JEDM | Journal of Educational Data Mining. 2020. Т. 12. № 3. С. 31–54. doi: 10.5281/zenodo.4143614.Gervet T. et al. When is Deep Learning the Best Approach to Knowledge Tracing? / T. Gervet // JEDM | Journal of Educational Data Mining. 2020. Т. 12. № 3. С. 31–54. doi: 10.5281/zenodo.4143614.Wilson K. H. et al. Back to the basics: Bayesian extensions of IRT outperform neural networks for proficiency estimation // arXiv preprint arXiv:1604.02336. 2016.Wilson K. H. et al. Back to the basics: Bayesian extensions of IRT outperform neural networks for proficiency estimation // arXiv preprint arXiv:1604.02336. 2016.Lindsey R. V. et al. Improving students’ long-term knowledge retention through personalized review / R. V. Lindsey // Psychological science. 2014. Т. 25. № 3. С. 639–647. doi:10.1177/0956797613504302.Lindsey R. V. et al. Improving students’ long-term knowledge retention through personalized review / R. V. Lindsey // Psychological science. 2014. Т. 25. № 3. С. 639–647. doi:10.1177/0956797613504302.Khajah M. M. et al. Integrating knowledge tracing and item response theory: A tale of two frameworks // CEUR Workshop proceedings. University of Pittsburgh, 2014. Т. 1181. С. 7–15.Khajah M. M. et al. Integrating knowledge tracing and item response theory: A tale of two frameworks // CEUR Workshop proceedings. University of Pittsburgh, 2014. Т. 1181. С. 7–15.Galyardt A., Goldin I. Move Your Lamp Post: Recent Data Reflects Learner Knowledge Better than Older Data / A. Galyardt, I. Goldin // Journal of Educational Data Mining. 2015. Т. 7. № 2. С. 83–108. doi: 10.5281/zenodo.3554671.Galyardt A., Goldin I. Move Your Lamp Post: Recent Data Reflects Learner Knowledge Better than Older Data / A. Galyardt, I. Goldin // Journal of Educational Data Mining. 2015. Т. 7. № 2. С. 83–108. doi: 10.5281/zenodo.3554671.Lan A. S., Studer C., Baraniuk R. G. Time-varying learning and content analytics via sparse factor analysis // Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. 2014. С. 452–461. doi: 10.1145/2623330.2623631.Lan A. S., Studer C., Baraniuk R. G. Time-varying learning and content analytics via sparse factor analysis // Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. 2014. С. 452–461. doi: 10.1145/2623330.2623631.Yudelson M. V., Koedinger K. R., Gordon G. J. Individualized Bayesian knowledge tracing models // International conference on artificial intelligence in education. Springer, Berlin, Heidelberg, 2013. С. 171–180.Yudelson M. V., Koedinger K. R., Gordon G. J. Individualized Bayesian knowledge tracing models // International conference on artificial intelligence in education. Springer, Berlin, Heidelberg, 2013. С. 171–180.Polson M. C., Richardson J. J. Foundations of intelligent tutoring systems. Psychology Press, 2013.Polson M. C., Richardson J. J. Foundations of intelligent tutoring systems. Psychology Press, 2013.Santoro A. et al. Meta-learning with memory-augmented neural networks // International conference on machine learning. PMLR, 2016. С. 1842–1850.Santoro A. et al. Meta-learning with memory-augmented neural networks // International conference on machine learning. PMLR, 2016. С. 1842–1850.

The authors declare that there are no conflicts of interest present.