The modern digital ecosystem is characterized by an unprecedented volume of information and a lightning-fast speed of its dissemination, which renders the forecasting of trends in social networks the cornerstone of effective strategic communications. Within the scope of the study, a critical-systemic evaluation of methodological tools of artificial intelligence (AI) employed for predictive analytics in social media was conducted, followed by their adaptation to the tasks of public relations specialists. The aim of the research is to develop a conceptual framework that integrates various algorithmic AI approaches—from natural language processing (NLP) and computer vision to network analysis—for constructing proactive strategies in corporate reputation management and communication campaigns. The empirical foundation of the study is formed through a systematic review of leading scientific publications dedicated to the described technologies and their applied capabilities in the context of detecting and predicting public sentiment. The scientific novelty of the work lies in the formulation of a unified methodological paradigm that shifts the emphasis of PR activities from passive response to informational challenges to active shaping of media discourse. The conclusions obtained will be useful both to researchers in the fields of communications, data processing, and computational linguistics, and to practitioners — PR directors and heads of digital agencies.

The rise of serverless and event-driven architectures is changing how we design and implement distributed systems. These approaches eliminate the need for server management and use event-based execution. They create systems that are more scalable, easy to deploy, resilient, and cost-effective. This paper looks at how serverless computing shifts the focus from managing infrastructure to focusing on application logic. At the same time, event-driven models improve responsiveness through asynchronous and independent communication. Together, they enhance the design of distributed systems by increasing fault tolerance, lowering operational costs, and supporting real-time, data-heavy applications. The discussion covers architectural principles, design choices, and new best practices that help integrate serverless and event-driven methods in current distributed computing. Ultimately, this study shows how these approaches lead to next-generation cloud-native systems that are flexible, quick, and tailored for changing workloads.

Background: The development of dexterous and intuitive prosthetic hands remains a significant challenge in rehabilitation engineering. Myoelectric control systems, which interpret surface electromyography (sEMG) signals from residual muscles, offer a promising avenue for non-invasive human-machine interfacing. However, traditional systems often suffer from limited accuracy, slow response times, and a lack of robustness to real-world conditions, hindering their clinical viability and user adoption. This article provides a comprehensive review of recent advancements aimed at overcoming these limitations.

Methods: We conducted a systematic review of contemporary literature focused on sEMG-based prosthetic control. The analysis covers the full spectrum of the control pipeline, including signal acquisition, pre-processing, and, most critically, feature extraction and pattern recognition. A special emphasis is placed on the transition from traditional machine learning classifiers to advanced deep learning architectures, particularly Convolutional Neural Networks (CNNs), for decoding hand gestures. The performance of these models is evaluated based on key metrics such as classification accuracy, computational latency, and robustness.

Results: The synthesis of recent findings reveals a clear trend: deep learning models, especially CNNs, consistently outperform traditional methods in hand gesture recognition accuracy, often exceeding 95% in controlled settings. Studies demonstrate that CNNs can automatically learn discriminative features from raw or minimally processed sEMG signals, eliminating the need for complex manual feature engineering. Furthermore, hybrid models and optimized network architectures have shown significant progress in achieving the low latency required for real-time prosthetic control.

Conclusion: Advanced signal analysis, powered by deep learning, represents a paradigm shift in myoelectric prosthetic control. These techniques are paving the way for more natural, reliable, and dexterous artificial limbs. Despite this progress, challenges related to inter-user variability, long-term stability, and clinical translation remain. Future research should focus on developing more generalizable models, integrating sensory feedback, and conducting extensive real-world usability studies to bridge the gap between laboratory breakthroughs and practical application.