KHRYSTYNA YURKIV

Ms. KHRYSTYNA YURKIV

orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
29 June 2023 in Measuring and computing devices in technological processes
Verified
Journal Article
Open access
PDF file
МКМирослав Комар
Мирослав Комар
HLHrystyna Ukrainian National Lipianina-Honcharenko
Hrystyna Ukrainian National Lipianina-Honcharenko
ІКІван Кіт
Іван Кіт
РМРоман Мадараш
Роман Мадараш
KYKHRYSTYNA YURKIV
KHRYSTYNA YURKIV
У сучасному світі проблема дезінформації набуває особливої актуальності, адже фейкові новини можуть мати серйозний вплив на громадську думку та політичні процеси. У даній статті представлено розроблений інтелектуальний метод виявлення джерел мультилінгвальної дезінформації, що публікуються в інтернеті. Метод базується на застосуванні алгоритмів машинного навчання та технологій обробки природної мови для аналізу тексту, векторизації, визначення схожості текстів та класифікації. Особливістю методу є його здатність працювати в режимі реального часу, адаптуватися до нових форматів дезінформації та проводити аналіз текстів на різних мовах. В статті детально описані основні етапи розробленого методу, його структура та порівняння з існуючими аналогами. Результати роботи можуть бути корисними для вчених, що займаються проблемами виявлення дезінформації, а також для розробників систем моніторингу та аналізу контенту в соціальних мережах та інших платформах.

Verified
Journal Article
Open access
PDF file
HLHrystyna Ukrainian National Lipianina-Honcharenko
Hrystyna Ukrainian National Lipianina-Honcharenko
KYKHRYSTYNA YURKIV
KHRYSTYNA YURKIV
Historical digital personas are a promising means of interactive and personalized access to cultural heritage; however, their practical deployment requires simultaneous control over factual accuracy, source provenance, persona consistency, and multimodal response latency. This paper proposes an evidence-grounded multimodal artificial intelligence (AI) framework for historical digital personas that combines controlled canonical interactions with open-ended retrieval-augmented generation. The architecture implements a cascaded query-processing strategy: high-confidence queries are routed to source-linked canonical frequently asked question (FAQ) responses and pre-generated multimedia objects, while open-ended queries are handled via hybrid full-text retrieval, cross-encoder reranking, an Evidence Gate, evidence-constrained generation, persona-oriented post-generation validation, speech synthesis, and lip synchronization. The framework is implemented and evaluated experimentally using the digital persona of Taras Shevchenko based on a Ukrainian-language cultural heritage knowledge base containing 543 evidence-oriented retrieval chunks, 50 canonical FAQ intents, and 500 natural language question variants. BGE-M3 hybrid retrieval achieved a Recall@1 of 36.46% and a Recall@10 of 77.71%, outperforming classical BM25 and compact dense retrieval baselines. Chunk-level cross-encoder reranking increased Recall@10 to 87.92% and normalized discounted cumulative gain at 10 (nDCG@10) to 0.6151. In a strict Leave-One-Variant-Out evaluation, the conservative FAQ cascade reached a coverage of 87.4% and a correctness of 95.19% among accepted matches. The Evidence Gate correctly rejected 96% of unsupported queries while retaining 88% of evidence-supported queries, achieving a Balanced Accuracy of 92%. Canonical multimedia caching reduced median processing time from approximately 143.94 s for complete dynamic multimedia executions to approximately 2.08 ms for processing-layer access to prepared canonical media under the evaluated warm-cache configuration. The results show that robust construction of historical digital personas requires a functional separation between high-confidence canonical interactions and open-ended, evidence-grounded generation, as well as coordinated yet functionally decoupled operation of retrieval, provenance, persona control, and multimedia delivery.

31 December 2023 in Herald of Khmelnytskyi National University. Technical sciences
Verified
Journal Article
Open access
PDF file
HLHrystyna Ukrainian National Lipianina-Honcharenko
Hrystyna Ukrainian National Lipianina-Honcharenko
МКМирослав Комар
Мирослав Комар
KYKHRYSTYNA YURKIV
KHRYSTYNA YURKIV
ВЛВасиль Лук‘яНчук
Василь Лук‘яНчук
In the context of modern warfare, the relevance of technogenic safety becomes particularly crucial due to the increasing threat of technogenic disasters. This article is dedicated to studying the most current challenges related to technogenic safety during wartime and determining the necessity for the development of novel approaches for prevention and management of the consequences of such catastrophes. The research object encompasses a complex of issues associated with mitigating the aftermath of technogenic disasters, while the main goal is the development of a model for intelligent assessment of these consequences. The aim of the work lies in devising a conceptual model for the intelligent assessment of the aftermath of technogenic catastrophes. To achieve this goal, the following tasks were undertaken: data collection regarding technogenic disasters, analysis of this data, development of a model for predicting consequences, determination of influencing parameters, and formulation of recommendations. The obtained results underscore the importance of creating such a model for swift and effective determination of the danger level in case of technogenic catastrophes. The conceptual model is built on the foundation of data collection, analysis, consequences modeling, and assessment, which allows for the consideration of diverse aspects of catastrophes and their impact on the environment and human beings. Evaluation results of the model demonstrated its effectiveness in classification and forecasting of danger levels. The achieved accuracy surpasses 77%, thus confirming the success of the approach. In the future, the model is intended to be expanded by considering additional factors, such as weather conditions and sensor readings, and applied across various fields of activity where operational assessment of danger levels is crucial. In conclusion, the article is dedicated to the development of a conceptual model for intelligent assessment of the consequences of technogenic catastrophes, which utilizes data, analysis, and modeling to swiftly and accurately determine danger levels, thereby enhancing safety in emergency situations.

10 September 2025 in RADIOELECTRONIC AND COMPUTER SYSTEMS
Verified
Journal Article
Closed access
PDF file
HLHrystyna Ukrainian National Lipianina-Honcharenko
Hrystyna Ukrainian National Lipianina-Honcharenko
MKMyroslav Komar
Myroslav Komar
HBHennadii Bohuta
Hennadii Bohuta
IIIhor Ihnatiev
Ihor Ihnatiev
KYKHRYSTYNA YURKIV
KHRYSTYNA YURKIV
LBLesia Bilovus
Lesia Bilovus
The subject matter is a general set of methods and system architecture for text analytics, enabling real-time detection and monitoring of information threats, validated through a Ukrainian case study. It integrates sentiment analysis, polarity-inversion handling, and machine-learning–based thematic classification. The research is especially relevant in the context of hybrid warfare, where the information environment becomes a battlefield of disinformation, manipulative campaigns and cognitive influence. The goal is to develop and experimentally validate a comprehensive information technology system for automated threat detection in the Ukrainian information space, built on the principles of Responsible Artificial Intelligence (Responsible AI) and modern natural language processing techniques. The objectives: the formation of a multilingual corpus of news and social media texts; implementation of a sentiment analysis module that incorporates polarity inversion; development of a hybrid thematic classification method that combines keyword dictionaries with machine learning model ensembles; and the construction of a Responsible AI Evaluation (RAIE) framework with indicators for fairness, transparency and user satisfaction. The obtained results confirm all five proposed hypotheses: the developed sentime оьnt analysis module achieves macro-F1 = 0.85 and reduces MAE by 18.2% compared to the baseline model; the polarity inversion detection algorithm allows automatic reversal of sentiment score in manipulative texts, improving the detection of hostile narratives; the hybrid thematic classification achieves macro-F1=0.83, with latency of 55 ms/document and throughput of 18 documents/second; integration of all modules into a unified pipeline improves recall by 10.4% without significant increase in latency; the RAIE conceptual model ensures ΔF1 ≤ 5%, an expert user satisfaction score of 4.14/5 and less than 10% latency overhead. The conclusions demonstrate that the proposed system effectively combines high accuracy in identifying information threats with the principles of ethical AI, transparency and user trust, making it practically valuable for national cybersecurity centres, CERTs and OSINT platforms. Conclusions. The scientific novelty lies in the development of novel methods: a context-sensitive sentiment analysis approach tailored to military-related vocabulary; a polarity inversion algorithm for detecting covert hostility; a hybrid thematic classification model combining machine learning with expert dictionaries; an integrated information processing architecture with >17 documents/second throughput; a Responsible AI evaluation model incorporating Fairness Gap, Model Cards and User Satisfaction Score.

8 October 2026 in Preprints.org
Verified
Preprint
Open access
PDF file
HLHrystyna Ukrainian National Lipianina-Honcharenko
Hrystyna Ukrainian National Lipianina-Honcharenko
KYKHRYSTYNA YURKIV
KHRYSTYNA YURKIV
Historical digital personas are a promising means of interactive and personalized access to cultural heritage; however, their practical deployment requires simultaneous control over factual accuracy, source provenance, persona consistency, and multimodal response latency. This paper proposes an evidence-grounded multimodal AI framework for historical digital personas that combines controlled canonical interactions with open-ended retrieval-augmented generation. The architecture implements a cascaded query processing strategy: high-confidence queries are routed to source-linked canonical FAQ responses and pre-generated multimedia objects, whereas open-ended queries are handled via hybrid full-text retrieval, cross-encoder reranking, an Evidence Gate, evidence-constrained generation, persona-oriented post-generation validation, speech synthesis, and lip synchronization. The framework is implemented and experimentally evaluated using the digital persona of Taras Shevchenko based on a Ukrainian-language cultural heritage knowledge base containing 543 evidence-oriented retrieval chunks, 50 canonical FAQ intents, and 500 natural language question variants. BGE-M3 hybrid retrieval achieved a Recall@1 of 36.46% and a Recall@10 of 77.71%, outperforming classical BM25 and compact dense retrieval baselines. Chunk-level cross-encoder reranking increased Recall@10 to 87.92% and nDCG@10 to 0.6151. In a strict Leave-One-Variant-Out evaluation, the conservative FAQ cascade reached a coverage of 87.4% and a correctness of 95.19% among accepted matches. The Evidence Gate correctly rejected 96% of unsupported queries while retaining 88% of evidence-supported queries, achieving a Balanced Accuracy of 92%. Canonical multimedia caching reduced median processing time from approximately 143.94 s for complete dynamic multimedia executions to approximately 2.08 ms for processing-layer access to prepared canonical media under the evaluated warm-cache configuration. The obtained results demonstrate that the robust construction of historical digital personas requires a functional separation of high-confidence canonical interactions and open-ended evidence-grounded generation, as well as a coordinated yet functionally decoupled operation of retrieval, provenance, persona control, and multimedia delivery.

28 May 2026 in Herald of Khmelnytskyi National University. Technical sciences
Verified
Journal Article
Open access
HLHrystyna Ukrainian National Lipianina-Honcharenko
Hrystyna Ukrainian National Lipianina-Honcharenko
МКМирослав Комар
Мирослав Комар
ПБПавло Биковий
Павло Биковий
KYKHRYSTYNA YURKIV
KHRYSTYNA YURKIV
The paper substantiates and develops an integration framework for the operational verification of responsible artificial intelligence, aligning international standards and framework documents of ISO/IEC, IEEE, OECD, UNESCO, NIST, and European Union regulatory approaches within a unified logic for assessing compliance of AI/ML systems. The study addresses the methodological gap between high-level ethical principles and their practical operationalization in measurable, reproducible, and auditable evaluation procedures. The proposed framework introduces a structured evidence chain: “principle – requirement – control/test – evidentiary artifact – managerial decision – post-market monitoring.” This logic ensures traceability, comparability, and auditability of assessment results across different types of AI systems, including traditional machine learning models, computer vision systems, natural language processing models, and generative AI. The core trustworthiness characteristics are systematized as robustness and safety, cyber resilience, privacy and data governance, transparency and explainability, fairness and non-discrimination, accountability, and traceability throughout the entire AI lifecycle. The complementary institutional roles are clarified: ISO/IEC establishes risk management and AI management system processes; the IEEE 7000 series operationalizes ethical and “by design” engineering requirements; OECD and UNESCO define value-based and human-rights-oriented principles; NIST provides a risk-oriented governance structure; and the European Union translates these principles into regulatory obligations, particularly for high-risk AI systems. The framework incorporates risk-based profiling of evaluation procedures and defines a standardized evidence package, including model and data documentation, testing protocols, risk registers and mitigation plans, operational logs, human oversight requirements, and continuous monitoring mechanisms. As a practical contribution, the study outlines the architecture and minimum functional components of a national/interagency responsible AI testing platform compatible with European risk-based compliance logic. The proposed approach transforms responsible AI from a declarative concept into an operational, verifiable, and auditable governance practice that strengthens regulatory alignment and public trust in AI-driven decision-making.