Due to the COVID-19 situation, all activities, including education, were shifted to online platforms. Consequently, instructors encountered increased challenges in evaluating students. In traditional assessment methods, instructors often face ambiguous cases when evaluating students’ competencies. Recent research has focused on the effectiveness of fuzzy logic in assessing students’ competencies, considering the presence of uncertain factors or multiple variables. Additionally, demographic characteristics, which can potentially influence students’ performance, are not typically utilized as inputs in the fuzzy logic method. Therefore, analyzing students’ performance by incorporating these factors is crucial in suggesting adjustments to teaching and learning strategies. In this study, we employ a combination of fuzzy logic and hierarchical linear regression to analyze students’ performance. The experiment involved 318 students from various programs and showed that the hybrid approach assessed students’ performance with greater nuance and adaptability when compared to a traditional method. Moreover, the findings in this study revealed the following: 1) There are differences in students’ performance between traditional and fuzzy evaluation methods; 2) The learning method is an impact on students’ fuzzy grades; 3) Students studying online do not perform better than those studying onsite. These findings suggest that instructors and educators should explore effective strategies being fair and suitable in assessment and learning.

Fake news detection has become a significant research top in natural language processing. Since the outbreak of the covid-19 epidemic, a large amount of fake news about covid-19 has spread on social media, making the detection of fake news a challenging task. Applying deep learning models may improve predictions. However, their lack of explainability poses a challenge to their widespread adoption and use in practical applications. This work aims to design a deep learning framework for accurate and explainable prediction of covid-19 fake news. First, we choose BiLSTM as the base model and improve the classification performance of the BiLSTM model by incorporating BERT-based distillation. Then, a post-hoc interpretation method SHAP is used to explain the classification results of the model to improve the transparency of the model and increase people's confidence in the practical application. Finally, utilizing visual interpretation methods, such as significance plots, to analyze specific sample classification results for gaining insights into the key terms that influence the model’s decisions. Ablation experiments demonstrated the reliability of the explainable method.

The article develops technology for generating song lyrics extensions using large language models, in particular the T5 model, to speed up, supplement, and increase the flexibility of the process of writing lyrics to songs with/without taking into account the style of a particular author. To create the data, 10 different artists were selected, and then their lyrics were selected. A total of 626 unique songs were obtained. After splitting each song into several pairs of input-output tapes, 1874 training instances and 465 test instances were obtained. Two language models, NSA and SA, were retrained for the task of generating song lyrics. For both models, t5-base was chosen as the base model. This version of T5 contains 223 million parameters. The analysis of the original data showed that the NSA model has less degraded results, and for the SA model, it is necessary to balance the amount of text for each author. Several text metrics such as BLEU, RougeL, and RougeN were calculated to quantitatively compare the results of the models and generation strategies. The value of the BLEU metric is the most diverse, and its value varies significantly depending on the strategy. At the same time, Rouge metrics have less variability and a smaller range of values. In total, for comparison, we used 8 different decoding methods for text generation supported by the transformers library, including Greedy search, Beam search, Diverse beam search, Multinomial sampling, Beam-search multinomial sampling, Top-k sampling, Top-p sampling, and Contrastive search. All the results of the lyrics comparison show that the best method for generating lyrics is beam search and its variations, including ray sampling. The contrastive search usually outperformed the usual greedy approach. The top-p and top-k methods do not have a clear advantage over each other, and in different situations, they produced different results.

Digital technologies and innovative methods have shown a significant impact on educational systems, and have made work easier for both learners and teachers. Additionally, they have improved the quality and the capability to digitize the assessment of student work produced during a learning process. Assessing and scoring students’ UML diagrams has become a challenging task for teachers, especially with the growing number of students, as well as the necessity to better manage their time. Consequently, there will be a necessity to automate the assessment of these learners. This paper presents an approach for assessing and grading automatically the student’s UML diagrams. The approach uses an algorithm implemented in Java, which takes the tutor's and student's solution diagrams as input, then provides the student's scores and identifies differences and errors made. Our algorithm was tested and evaluated in a real case within a web platform, and the results obtained demonstrate the effectiveness of our solution.

Numerous facets of life are impacted by the efficient application of technologies. Education, like all other fields, is a major area where technology is used to teach and learn effectively. One of these technologies that instructors and educators have recently become interested in is blended learning. This article aims at identifying the main constructs that highly influence the adoption of blended learning in higher education through meta-analytic literature review and proposing new technology acceptance model that is suitable for digital education tools. About 32 quantitative studies published since 2007 in journals and conferences are selected for performing weight computation and meta-analysis of constructs with a total sample size of 8,168. Moreover, the study also conceptualises the new technology acceptance model for digital education tools, considering both students and instructors as end users. The descriptive statistics indicate that there has been an increase in the number of publications since the year 2020. The results show that perceived ease of use, attitude toward usage, and perceived usefulness on intention to use are good predictors concerning student respondents, while performance expectancy on behavioural intention is found to be a good predictor concerning instructors. The results of the meta-analysis highlight the significance of blended learning in government and educational institutions that prioritises student ease of use and instructor performance expectancy and facilitating conditions. The results prove the effectiveness of blended learning in enhancing student learning experiences while improving educational practises and outcomes. In addition to identifying key factors influencing blended learning adoption in higher education, the study also suggests a novel model for implementing digital education tools, considering both student and instructor perspectives, thereby addressing a critical need in educational technology research. It offers actionable recommendations for policymakers and educational institutions to enhance the quality of higher learning.

Web-based learning systems have quickly developed, by giving students a broader access to wide range of courses. However, when presented with a huge number of courses, it might be difficult for users to rapidly discover the ones they are interested in, from a large amount of online educational resources. As a result, a course recommendation system is crucial to increase users' learning benefit. Presently, numerous online learning platforms have developed a variety of recommender systems using conventional data mining techniques. Still, these methods have several shortcomings, like adaptability and sparsity. To solve this problem, this study provides a deep learning based English course recommendation system with the extraction of features using a dual channel based capsule network (CapsNet). This network extracts all the important features about the courses and learners and suggests suitable courses for the learners. To evaluate the proposed model’s performance, several investigations are performed on a real-world dataset (XuetangX) and outperforms existing recommendation approaches with an average of 91% precision, 45% recall, 55% f1-score, 0.798 RMSE, and 0.671 MSE. According to the experimental findings, the proposed model provides better and more reliable recommendation performance than the conventional approaches. According to the experimental findings, the proposed model provides better and more reliable recommendation performance than the conventional approaches.

Assessing pre-service teachers’ digital literacy is challenging, particularly in inclusive education. Reliable and valid testing instruments are required to measure the digital literacy pre-service teachers possess in inclusive education. The entire research process comprises three phases. The first stage was to develop the assessment instrument, the second stage was to validate its content validity, and a pilot study was then conducted to test the reliability and construct validity of the instrument. The results of this study showed that item-level and scale-level content validity scores were both 1.0. The Kaiser-Meyer-Olkin is equal to 0.865. Five factors were extracted, explaining 54.40% of the total variance. The model fits were also all satisfactory. Standardized factor loadings of the instrument’ s 28 items were above 0.5. The values of Cronbach’s are higher than 0.7 for the five factors and the whole instrument. It can be summarized that the instrument had good reliability and validity and can be used to assess the digital literacy of pre-service teachers in inclusive education. There has been research into developing tools to evaluate the digital literacy of pre-service teachers. Still, few studies have addressed pre-service teachers of inclusive education, and this study fills this research gap. The subsequent phase involves evaluating it using a more extensive sample.