This paper presents a multi-level annotation model for detecting fake news in Kazakh and Russian languages, aiming to enhance understanding of disinformation strategies in multilingual digital media environments. Unlike traditional binary models, our approach captures the complexity of disinformation by accounting for both linguistic and cultural factors. To support this, a corpus of over 5000 news texts was manually annotated using the Label Studio platform. The annotation scheme consists of seven interrelated categories: CLAIM, SOURCE, EVIDENCE, DISINFORMATION_TECHNIQUE, AUTHOR_INTENT, TARGET_AUDIENCE, and TIMESTAMP. Inter-annotator agreement, evaluated using Cohen’s Kappa, ranged from 0.72 to 0.81, indicating substantial consistency. The annotated data reveals recurring patterns of disinformation, such as emotional manipulation, targeting of vulnerable individuals, and the strategic concealment of intent. Semantic relations between entities, such as CLAIM → EVIDENCE and CLAIM → AUTHOR_INTENT were formalized to represent disinformation narratives as knowledge graphs. This study contributes the first linguistically and culturally adapted annotation model for Kazakh and Russian languages, providing a robust and empirical resource for building interpretable and context-aware fake news detection systems. The resulting annotated corpus and its semantic structure offer valuable empirical material for further research in natural language processing, computational linguistics, and media studies in low-resource language environments.

This paper presents a comprehensive study aimed at systematically analyzing and evaluating natural language processing (NLP) techniques for military information operations, with a special focus on social media intelligence. Among an ever-growing complicated information environment, NLP methods like sentiment analysis, named entity recognition, and topic modeling have been essential in tracking online propaganda efforts, discovering emerging issues and threats globally with dialogues on military operations. These techniques make an impact on available decision making via situational awareness and getting the added extraction from volumes of unstructured data outputs thus increasing the overall strategic benefits to military organizations. There are technical and operational challenges concerning the use of NLP in a military context such as requirements for real-time data processing; language diversity; and maintaining data privacy while preserving ethical standards. To address these challenges, the study conducts an exhaustive survey of NLP methods, reviewing their range of applications, and highlights the relevance of several approaches for military information operations, with special emphasis on social media intelligence. The work further provides discussion on the comprehensive adoption of artificial intelligence (AI), edge computing, and multilingual NLP models for enhancing adaptability, efficiency, and transparency of the systems. It also extols the need for explainable AI (XAI) to improve accountability and trust by making term or even whole early warning systems derived from NLP analyses, transparent and interpretable for these military research applications with significant financial consequences. The paper also emphasizes the strategic importance of multilingual and multimodal analysis and the integration of specialized military lexicons to improve the contextual understanding of military discourse in social media environments. We also elucidate the important capabilities of NLP in enabling military operations to be responsive, rapid and data-driven while also adapting to the evolving nature of warfare. Key conclusions suggest that applying advanced NLP tools enhances situational awareness, enables timely threat detection, and supports more agile, data-informed decision-making within modern military operations. The paper shows a perspective to optimize NLP and AI technologies, leveraging various perspectives to benefit the operational needs of military and defense sectors in more data-rich environments.

Multimodal human–AI systems generally consider facial expressions and body motions as separate input streams, leading to disjointed interpretations and diminished emotional coherence. To overcome this issue, we offer the Engagement-Safe Expressive Alignment (ESEA) paradigm and the Unified Visual Synchrony (UVS) framework as its computational implementation. UVS models the coherence between facial expressions and gestures, offering an interpretable visual synchrony signal that can function as adaptive feedback in human–AI interactions. The framework’s key component is the Consistency Index for Affective Synchrony ($CIAS$), which correlates brief visual segments with scalar synchrony scores through a common latent representation. Facial and gestural signals are processed by modality-specific projection networks into a unified latent space, and $CIAS$ is derived from the similarity and short-term temporal consistency of these latent trajectories. The synchrony index is regarded as an estimation of affective visual coherence within the ESEA paradigm. We formalize the UVS/$CIAS$ framework and conduct a comparative experimental evaluation utilizing matched and mismatched face–gesture segments derived from rendered dialog footage. Utilizing ROC analysis, score distribution comparisons, temporal visualizations, and negative control tests, we illustrate that $CIAS$ effectively captures structured face–gesture alignment that surpasses similarity-based baselines, while also delivering a persistent, time-resolved synchronization signal. These findings establish $CIAS$ as a principled and interpretable feedback signal for future affect-aware, engagement-focused multimodal agents.

As false news benchmarks are increasingly being developed from fact-checking sources, this introduces a slight, but under cited, threat to validity: source bias. The two classes are not necessarily similar in style or structure; a classifier could distinguish between fake and real items (both from a fact-checking outlet) based on surface features rather than truth, and obtain scores which do not reflect the ability of the classifier to actually detect the truth. We report this effect for Kazakh–Russian Telegram—a low-resource code-switched environment—and show that a naively engineered corpus yields artificially high performance (up to 0.98 F1) that plummets by about ten percentage points, after the removal of the shortcut. We then introduce a two-stage debiasing protocol: (1) structural cleaning (viral-boilerplate, interrogative openers and verdict labels removal from Tengrinews posts); and (2) a heterogeneous two-source real class constructed from Tengrinews posts and confirmed “True” (Pravda) verdicts on Factcheck.kz. This balanced, cleaned set of 804 posts (402 real/402 fake) is made available to the public. On this debiased benchmark we report honest baselines for six methods: TF-IDF + Logistic Regression, FastText, fine-tuned XLM-RoBERTa, fine-tuned RuBERT, and Llama 3.1 8B with zero- and few-shot prompting. The fine-tuned transformers show good performance (RuBERT 0.8819, XLM-RoBERTa 0.8757 macro F1), whereas classical TF-IDF is competitive (0.8508), and in-context LLM prompting lags behind (zero-shot 0.4522, few-shot 0.6601 macro F1)–quantifying the gap between in-context prompting and few hundred labeled examples. The transformer margin is very small and is calculated from a single split, so we only report the margin as indicative and discuss the evaluation conditions required to confirm the margin. We make a methodological contribution by providing a replicable debiasing recipe and showing that explicit source-bias controls are required to evaluate trustworthy fact-checks in low-resource, fact-check-derived misinformation benchmarks. Data and code are made publicly available so that they can be used to facilitate reproducible research.

В работе сравниваются лексическая, моноязычная Transformer-модель и экспериментальная модель объединения вероятностных оценок на очищенных казахскоязычных заголовках и кратких описаниях Factcheck.kz. После удаления явных формулировок вердиктов и шаблонных фраз источника 1 265 записей были хронологически разделены на 885 обучающих, 190 валидационных и 190 тестовых. Операциональный положительный класс объединяет категории «ложь», «фейк», «манипуляция», «полуправда» и «непроверено» и поэтому не представляет собой однородный класс полностью вымышленных сообщений. Модель TF-IDF + Linear SVM достигла macro-F1 = 0,6563 при широком 95%-ном bootstrap-доверительном интервале [0,5834; 0,7253]. Transformer-модель показала среднее значение macro-F1 = 0,6585 (SD = 0,0442) по пяти random seeds. Экспериментальная модель score fusion получила 0,6422 (SD = 0,0180) и не выполнила заранее установленный критерий превосходства над обоими компонентными моделями по среднему значению test macro-F1. Парные bootstrap-интервалы различий между тремя подходами включали ноль, поэтому статистически подтвержденное превосходство какой-либо модели не установлено. В дополнительном анализе положительный класс был ограничен категориями «фейк» и «манипуляция». Переобученная SVM достигла macro-F1 = 0,7153 на 123 тестовых записях, тогда как исходная модель с широким набором меток на той же подвыборке показала 0,7488. Следовательно, преимущество переобучения на более узком наборе категорий не подтверждено. Оценка ограничена одним редакционным источником: хронологическое разделение и очистка текста сами по себе не подтверждают переносимость результатов на другие публикационные источники или исходные утверждения. Полученные результаты указывают на необходимость осторожной интерпретации результатов классификации на малых выборках. Использование модели в качестве инструмента предварительного отбора материалов требует независимой внешней валидации и экспертного контроля.