Кафедра штучного інтелекту (ШІ)
Постійний URI для цієї колекції
Перегляд
Останні подання
Публікація Дослідження методів аналізу емоційного забарвлення текстів(ХНУРЕ, 2025) Боков, І. П.; Петров, К. Е.This work is devoted to the study of methods for determining the emotional coloring of texts. Traditional approaches, such as dictionary-based and rule-based methods, as well as modern machine learning techniques, including neural networks and transformer models were analyzed. Special attention is given to the challenges of sentiment analysis, such as sarcasm, irony, and contextual word meanings. The study highlights the advantages and limitations of different approaches and emphasizes the need for optimizing transformer models for resource-limited environments.Публікація Застосування моделей NLP для оптимізації пошуку в файлових сховищах(ХНУРЕ, 2025) Коваль, Г. К.; Гриньова, О. Є.This research implements AI-powered search for file storage platforms using vector search and NLP models. The system analyzes and indexes documents, enabling semantic searches based on content and context. It uses Sentence Transformers for vector representations and Pinecone for efficient storage and retrieval. This innovation overcomes traditional search limitations, enhancing document management efficiency across various domains. The project aims to improve user productivity by providing more relevant search results and streamlining file retrieval processes in large-scale document repositories.Публікація Створення та корекція зображення за допомогою текстових підказок у діалозі(ХНУРЕ, 2025) Левицький, К. Ю.; Терзіян, В. Я.Generative AI models for image creation use text prompts by encoding them into numerical representations with transformers and applying diffusion models to refine noisy images. The process involves multiple convolution operations to gradually enhance image quality. These models understand not only keywords but also contextual relationships, allowing for complex scene generation. To enhance generated images, high-resolution upscaling methods reduce blurriness and ensure stylistic consistency. Further research is needed to improve precision, reduce variability, and optimize user control over AI-generated visuals.Публікація Проблеми інтеграції штучного інтелекту в хірургію(ХНУРЕ, 2025) Міхеєва, М. О.; Губін, В. О.The combination of artificial intelligence with medicine, namely with the direction of surgery and robotic surgery, allows you to improve the process of performing operations and increase the accuracy of manipulations. The article discusses the advantages and disadvantages of integrating artificial intelligence into the field of medical surgery, and also discusses the interaction of AI with robotic systems. It also discusses the problems that may arise during this activity and ways to solve them.Публікація Методи оцінки якості синхронізації аудіо та відео на основі нейронних мереж(ХНУРЕ, 2025) Мирошник, Ю. Ю.; Рябова, Н. В.This thesis analyses the impact of audiovisual synchronisation on the realism of digital avatars used in film dubbing, video conferencing and digital assistants. Accurate lip-audio alignment is highlighted as a key factor in achieving natural, Comparative attention is given to specialised lip-sync models (e.g. SyncNet) and advanced multimodal approaches (e.g. AV-HuBERT). The results show that stable synchronisation significantly improves audience engagement and reduces visual inconsistencies. Practical applications in multilingual face-dubbing illustrate the potential and limitations of current methods.Публікація Системи з використанням LLM як мотиватор для навчання(ХНУРЕ, 2025) Небаба, М. Ю.; Рябова, Н. В.This work investigates the integration of artificial intelligence with multimedia animation in virtual streaming platforms. The system architecture is based on a large language model (LLM). Real-time text generation and adaptive linguistic modulation are achieved through advanced algorithmic techniques. Synchronized voice and animated avatar interactions are facilitated by efficient API-based connectivity. Reviewed Neuro-sama system, mentioned opensource platforms and APIs, and how they encourage learning and experimenting with machine learning and data science.Публікація Нейромережева модель прогнозування ціни ринкових активів для заданих часових фреймів(ХНУРЕ, 2025) Родіонов, І. О.; Чала, Л. Е.Modern computational methods and artificial intelligence play a crucial role in personal finance and financial market trading. The main challenge is developing intelligent systems capable of real-time data processing, risk assessment, and personalized financial recommendations. The relevance of this issue is driven by market volatility, increasing data complexity, and the need for automation in decision-making. Existing solutions include algorithmic trading, robo-advisors, and AI-driven risk management. However, these approaches face limitations in adaptability and transparency. Our proposed solution focuses on enhancing AI models with explainability, real-time analytics, and adaptive learning to improve financial decision-making.Публікація Моніторинг та відновлення сільськогосподарських земель засобами штучного інтелекту(ХНУРЕ, 2025) Сільванович, К. В.; Гриньова, О. Є.This research examines the analysis and classification of agricultural land damage using aerial photography, and also methods for predicting the time required for their restoration. Utilizing modern machine learning technologies, an architecture for an intelligent decision support system has been proposed to identify soil damage and predict recovery timelines. The system comprises three AI models. The results aim to enhance decision-making processes in the restoration of agricultural resources, promoting sustainable management of lands post-conflict.Публікація Дослідження методів аналізу руху людини за допомогою дронів(ХНУРЕ, 2025) Сагайдачний, Н. І.; Золотухін, О. В.In modern conditions, there is an increasing demand for high-precision monitoring systems capable of analyzing human movement in real time. One promising approach is the use of unmanned aerial vehicles (UAVs) equipped with cameras and computer vision algorithms. Such systems can be applied in security, sports analytics, medicine, and military operations.Публікація Дослідження методів та моделей управління IT-проєктами з розробки комп’ютерних ігор(ХНУРЕ, 2025) Дроздов, Я. Д.; Петров, К. Е.This paper examines existing models and methods of IT project management in the field of computer game development. Popular methodologies such as Agile, Scrum and Kanban are analyzed, as well as their application in the game development industry. The main problems and limitations of these approaches are identified. Based on the conducted research, an experimental method of game project management is developed, which takes into account the features of game development and combines elements of existing methodologies. The study proposes the partial probabilistic forecasting as the main method to use.Публікація Використання технологій штучного інтелекту для створення відео на основі тексту(ХНУРЕ, 2025) Шатило, І. Ю.; Чала, Л. Е.This paper introduces a technology designed for automated video generation from web-based articles. The system follows a structured process that includes content parsing, translating it, summarising key points, extracting keywords, and analysing sentiment. It selects relevant video clips and background music, synthesises voiceovers using text-to-speech technology, and assembles the final video. The result is a fully automated pipeline that converts textual content into dynamic video presentations, with applications in news automation, educational media, and digital content production.Публікація Концептуальна модель доповнення точкових хмар на основі графових нейронних мереж(ХНУРЕ, 2025) Чухран, І. Д.; Чала, Л. Е.A conceptual model for completing point clouds using graph neural networks has been developed, which allows to efficiently encode incomplete point clouds as graphs and predict missing points. The proposed approach allows for high-quality reconstruction of incomplete 3D data, which is extremely important for many applications such as 3D reconstruction, robot navigation, etc. Research methods are analysis and synthesis of scientific literature, systematic approach, machine learning methods, in particular graph neural networks. Further research may involve testing different neural network architectures, tuning hyperparameters, applying alternative loss functions, and using more powerful computing resources to train deep models.Публікація Застосування reinforcement learning для алгоритмічної торгівлі на фінансових ринках(ХНУРЕ, 2025) Черкасов, М. М.; Чала, Л. Е.This work explores the application of reinforcement learning for algorithmic trading in financial markets. It identifies the challenge of automating decision-making amid market volatility and uncertainty. The study reviews existing solutions such as Q-learning, deep Q-networks, and Actor-Critic architectures. It highlights the potential of hybrid models that integrate supervised pre-training with reinforcement learning. The proposed approach also incorporates risk management mechanisms to dynamically adapt trading strategies.Публікація Сучасні проблеми обчислювального та штучного інтелекту в сфері ведення особистих фінансів та торгівлі на фінансових ринках(ХНУРЕ, 2025) Черкасов, М. М.; Чала, Л. Е.Modern computational methods and artificial intelligence play a crucial role in personal finance and financial market trading. The main challenge is developing intelligent systems capable of real-time data processing, risk assessment, and personalized financial recommendations. The relevance of this issue is driven by market volatility, increasing data complexity, and the need for automation in decision-making. Existing solutions include algorithmic trading, robo-advisors, and AI-driven risk management. However, these approaches face limitations in adaptability and transparency. Our proposed solution focuses on enhancing AI models with explainability, real-time analytics, and adaptive learning to improve financial decision-making.Публікація Застосування методів глибого навчання в задачах NPL: граматична корекція текстів(ХНУРЕ, 2025) Харченко, М. В.; Рябова, Н. В.. Modern approaches to solving the problem of text Grammatical Error Correction (GEC), based on deep learning methods, are considered. The main algorithmic methods, including rule-based, dictionary-based and more recent neural network-based methods, are considered. The problem of regeneration of whole input sequence in sequence-to-sequence models is described. The sequence-to-edits method with an example of such architecture is described as a solution to the regeneration problem. A perspective approach to combining algorithmic and neural network methods is proposed.Публікація Огляд задачі інтелектуального аналізу тексту для розпізнавання дезінформації(ХНУРЕ, 2025) Харіна, О. С.; Головянко, М. В.This work discusses the challenges posed by the abundance of information available today, which facilitates the rapid spread of misinformation. The study focuses on developing machine learning algorithms for automated detection and classification of misinformation, addressing the importance of a combined approach that includes technological, social and psychological aspects. An objective and a plan have been developed to investigate machine learning tools for misinformation detection, taking into account the limitations of existing datasets.Публікація Енергоспоживання та продуктивність моделей ші(ХНУРЕ, 2025) Фесенко, М. Б.; Вітько, О. В.The increasing computational complexity of modern artificial intelligence (AI) models has led to a significant rise in energy consumption. The training and deployment of large-scale neural networks require substantial computational power, raising concerns about environmental sustainability. This paper examines energy efficiency issues in deep learning, the role of optical processors in AI acceleration, and the comparative performance of GPU and TPU architectures in training neural networks.Публікація Обмеженість сучасних моделей ші та потреба в загальному інтелекті(ХНУРЕ, 2025) Фесенко, М. Б.; Вітько, О. В.The current artificial intelligence (AI) systems achieve high performance in narrow tasks but lack generalization capabilities inherent to human intelligence. This limitation hinders adaptability and problem-solving in unfamiliar environments without additional training. This paper examines key shortcomings of modern AI models and justifies the need for artificial general intelligence (AGI) capable of flexible learning, logical reasoning, and autonomous adaptation.Публікація Підхід до створення інтелектуального помічника для наукових конференцій на основі LLM та PFE(ХНУРЕ, 2025) Теплюк, М. В.; Гриньова, О. Є.The article describes the development of an intelligent assistant for the website of the International Forum of Young Scientists «Radio Electronics and Youth in the XXI Century» to automate the search for similar scientific papers and answer frequently asked questions. A combination of NLU and PFE is used. The system is based on the SBERT model for fast semantic text comparison, Ukr- RoBERTa-base for Ukrainian language processing, and ChatGPT for checking the structure of papers. This solution will enhance the forum's efficiency and the quality of research.Публікація Challenges in named entity recognition(ХНУРЕ, 2025) Serhiienko, D. V.; Ryabova, N. V.This work explores the challenges and advancements in Named Entity Recognition (NER) within natural language processing. The study provides a comprehensive review of the key difficulties faced in NER. Additionally, it examines contemporary methods and models designed to address these challenges, highlighting recent innovations in deep learning and transformer-based architectures. The paper also investigates the impact of linguistic diversity, noisy data, and evolving language on NER performance, offering insights into optimization strategies and future research directions in entity recognition.