The paper proposes a solution to improve the efficiency of recognition of speech defects in children by processing the sound data of the spectrogram based on convolutional neural network models. For a successful existence in society, a person needs the most important skill - the ability to communicate with other people. The main part of the information a person transmits through speech. The normal development of children necessarily includes the mastery of coherent speech. Speech is not an innate skill for people, and children learn it on their own. Speech defects can cause the development of complexes in a child. Therefore, it is very important to eliminate them at an early age. So, the problem of determining speech defects in children today is a very urgent problem for parents, speech therapists and psychologists. Modern information technologies can help in solving this problem. The paper provides an analysis of the literature, which showed that models of CNN can be successfully used for this. But the results that are available today have not been applied to speech in Ukrainian. Therefore, it is important to develop and study models and methods of convolutional neural networks to identify violations in the speech of children. The paper describes a mathematical model of oral speech disorders in children, the structure of a convolutional neural network and the results of experiments. The results obtained in the work allow to establish one of the speech defects: dyslexia, stuttering, difsonia or dyslalia with recognition results of 77-79%.
Augmented reality is currently one of the most rapidly emerging technologies, with applications in the field of education. AR technology, according to a review of numerous publications, provides new options for teachers and makes learning more appealing to pupils of all ages. Students can use mobile AR apps to observe a real-world environment with virtual things layered or composited on top of it. This is especially true for young children. The article highlights the activities that allow a child to have a unique AR experience. There is a comparison of available augmented reality (AR) applications for learning the alphabet. The author describes a novel augmented reality application that was created with Unity, C\#, and Vuforia. The developed mobile AR application allows students to learn the Ukrainian alphabet, number names, and animal sounds. The learning process is enhanced by three-dimensional visualization and sounding of each letter and number thanks to this program. An analysis of a survey of teachers and parents found that utilizing an AR application greatly boosted children's motivation and self-efficacy in learning letters and numbers. The AR application aided in the retention of the child's attention while learning new content by speeding up the process of memorization.
In the modern conditions of digitalization of education, the issue of objective assessment of learning outcomes and timely identification of students who need additional support is of particular importance. This work considers the possibilities of using machine learning algorithms to predict the final results of semester control. The relevance of the study is due to the need for tools for objectifying assessment and timely identification of risks of failure to adjust the individual educational trajectory. The study is aimed not only at obtaining an accurate numerical score, but also at determining the level of student success in a generalized categorical form (from «Very low» to «Good»), which allows for a deeper analysis of the quality of the educational process and timely adjustment of the individual learning trajectory. The work uses the «Students Performance Factors» data set, which includes indicators of class attendance, intensity of study, previous academic training, and socio-economic factors. Data preparation included anomaly removal, feature normalization, and correlation analysis to select the most informative parameters. For the experimental study, linear regression, multilayer perceptron (neural network), decision trees, and random forest were used. The models were evaluated using R² and MAE metrics for regression tasks and accuracy for classification tasks. The results showed that linear regression was the most appropriate for predicting the exact final score, demonstrating high accuracy and ease of interpretation. At the same time, in the task of distributing students by achievement levels, the best results were obtained using a neural network. Decision tree analysis confirmed the important role of attendance as a key factor in academic success. The results obtained show that the choice of method should depend on the forecast format: for quantitative assessments, it is advisable to use regression, and for qualitative separation of students by levels, neural networks. The study opens up prospects for implementing educational analytics tools to improve learning efficiency
30 September 2023 in Ukrainian Journal of Educational Studies and Information Technology
At present, modern information technologies are developing very rapidly. A lot of different software tools are being created to improve and simplify people's lives. This is especially true in the context of distance learning. In the context of online learning, the traditional form of conducting dictations needs to be transformed. Therefore, it is possible to propose automation of the process of conducting and checking dictations using software. The paper analyzes scientific research and publications of the current state of language synthesis technologies and text similarity testing. With the help of C# and the language synthesis libraries of Microsoft, Google, Amazon, software has been developed that allows the user to listen and type text, and then automatically check it with the initial sample. The conducted testing for different types of errors showed the possibility and expediency of development the system for conducting and checking dictations. Follow-up activities will focus on improving the effectiveness of the program. The use of such programs will help improve the organization of conducting and checking dictations during distance learning and self-training of students.
Document Layout Analysis (DLA) is a critical step in intelligent document processing and is essential for accurately reconstructing the hierarchical structure of pages. While modern convolutional neural networks exhibit high performance, their effectiveness heavily depends on the quality and representativeness of training data, limiting their application in scenarios where labeled datasets are scarce. This paper proposes a method for enhancing DLA through synthetic generation of training data. A formalized mathematical model for generating document layouts has been developed, allowing control over element placement density, sizes, and spatial distribution. An experimental study investigated the impact of various data generation strategies on the training of the YOLO11m model, including median and threshold-based element splitting as well as different block sampling schemes. The experiments showed that employing median element splitting combined with random sampling from a large shuffled pool of synthetic data yields consistent improvements of 2–4% across all key metrics: precision, recall, mAP@50, and mAP@50:95, as compared with simple data generation strategies. These results demonstrate that targeted optimization of the data preparation process can enhance the performance of convolutional models in DLA tasks without increasing architectural complexity. The practical applicability of the method is validated through integration into the MinerU system. Future research will focus on extending the proposed model to complex layouts in scientific journals, technical reports, and handwritten documents.