Search Results (616)

Existing approaches to social media sentiment analysis typically focus on static classification, offering limited foresight into how public opinion evolves. This study addresses that gap by introducing the Multi-Feature Sentiment-Driven Forecasting (MFSF) framework, a novel pipeline that enhances sentiment trend prediction by integrating rich contextual information from text. Using state-of-the-art transformer models on the Sentiment140 dataset, our framework extracts three concurrent signals from each tweet: sentiment polarity, aspect-based scores (e.g., ‘price’ and ‘service’), and topic embeddings. These features are aggregated into a daily multivariate time series. We then employ a SARIMAX model to forecast future sentiment, using the extracted aspect and topic data as predictive exogenous variables. Our results, validated on the historical Sentiment140 Twitter dataset, demonstrate the framework’s superior performance. The proposed multivariate model achieved a 26.6% improvement in forecasting accuracy (RMSE) over a traditional univariate ARIMA baseline. The analysis confirmed that conversational aspects like ‘service’ and ‘quality’ are statistically significant predictors of future sentiment. By leveraging the contextual drivers of conversation, the MFSF framework provides a more accurate and interpretable tool for businesses and policymakers to proactively monitor and anticipate shifts in public opinion. Full article

Certain industries will bear a disproportionate share of the burden of climate change. Climate change risk perceptions can impact workers’ mental health and well-being; increased climate change risk perceptions are also associated with more favourable adaptive attitudes. It is, therefore, important to understand whether climate risk perceptions differ across workers between industries. We conducted an online survey of British Columbians (16+) in 2021 using social media advertisements. Participants rated how likely they believed their industry (Natural Resources, Science, Art and Recreation, Education/Law/Government, Health, Management/Business, Manufacturing, Sales, Trades) would be affected by climate change (on a scale from “Very Unlikely” to “Very Likely”). Ordinal logistic regression examined the association between occupational category and perceived industry vulnerability, adjusting for socio-demographic factors. Among 877 participants, 66.1% of Natural Resources workers perceived it was very/somewhat likely that climate change would impact their industry; only those in Science (78.3%) and Art and Recreation (71.4%) occupations had higher percentages. In the adjusted model, compared to Natural Resources workers, respondents in other occupations, including those in Art and Recreation, Education/Law/Government, Management/Business, Manufacturing, Sales, and Trades, perceived significantly lower risk of climate change-related industry impacts. Industry-specific interventions are needed to increase awareness of and readiness for climate adaptation. Policymakers and industry leaders should prioritize sectoral differences when designing interventions to support climate resilience in the workforce. Full article

Social media has become a platform for the rapid spread of persuasion techniques that can negatively affect individuals and society. Propaganda detection, a crucial task in natural language processing, aims to identify manipulative content in texts, particularly in news media, by assessing propagandistic intent. Although extensively studied in English, Arabic propaganda detection remains challenging because of the language’s morphological complexity and limited resources. Furthermore, most research has treated propaganda detection as an isolated task, neglecting the influence of sentiments and emotions. The current study addresses this gap by introducing the first multi-task learning (MTL) models for Arabic propaganda detection, integrating sentiment analysis and emotion detection as auxiliary tasks. Three MTL models are introduced: (1) MTL combining all tasks, (2) PSMTL (propaganda and sentiment), and (3) PEMTL (propaganda and emotion) based on transformer architectures. Additionally, seven task-weighting schemes are proposed and evaluated. Experiments demonstrated the superiority of our framework over state-of-the-art methods, achieving a Macro-F1 score of 0.778 and 79% accuracy. The results highlight the importance of integrating sentiment and emotion for enhanced propaganda detection; demonstrate that MTL improves model performance; and provide valuable insights into the interaction among sentiment, emotion, and propaganda. Full article

The vast volume of media content, combined with the costs of manual annotation, challenges scalable codebook analysis and risks reducing decision-making accuracy. This study evaluates the effectiveness of large language models (LLMs) and multi-agent teams in structured media content analysis based on codebook-driven annotation. We construct a dataset of 200 news articles on U.S. tariff policies, manually annotated using a 26-question codebook encompassing 122 distinct codes, to establish a rigorous ground truth. Seven state-of-the-art LLMs, spanning low- to high-capacity tiers, are assessed under a unified zero-shot prompting framework incorporating role-based instructions and schema-constrained outputs. Experimental results show weighted global F1-scores between 0.636 and 0.822, with Claude-3-7-Sonnet achieving the highest direct-prompt performance. To examine the potential of agentic orchestration, we propose and develop a multi-agent system using Meta’s Llama 4 Maverick, incorporating expert role profiling, shared memory, and coordinated planning. This architecture improves the overall F1-score over the direct prompting baseline from 0.757 to 0.805 and demonstrates consistent gains across binary, categorical, and multi-label tasks, approaching commercial-level accuracy while maintaining a favorable cost–performance profile. These findings highlight the viability of LLMs, both in direct and agentic configurations, for automating structured content analysis. Full article

Deepfake detection has become a critical issue due to the rise of synthetic media and its potential for misuse. In this paper, we propose a novel approach to deepfake detection by combining video frame analysis with facial microexpression features. The dual-branch fusion model utilizes a 3D ResNet18 for spatiotemporal feature extraction and a transformer model to capture microexpression patterns, which are difficult to replicate in manipulated content. We evaluate the model on the widely used FaceForensics++ (FF++) dataset and demonstrate that our approach outperforms existing state-of-the-art methods, achieving 99.81% accuracy and a perfect ROC-AUC score of 100%. The proposed method highlights the importance of integrating diverse data sources for deepfake detection, addressing some of the current limitations of existing systems. Full article

The presence of persistent contaminants in soils is of growing concern around the world. Contaminated soils can affect numerous ecological environments and lead to significant health risks to humans, affecting soil biodiversity, structure and geomechanical behaviour and agricultural sustainability. Additionally, soil contaminants can also leach into water flows, which is another concern. In general, soil contamination can be attributed to natural sources or to anthropogenic sources associated with human activity. Soil contaminants are usually classified in the following categories: biological, radioactive, organic and inorganic contaminants. State of the art information regarding some of the most common persistent soil contaminants, including possible sources and prevalence, and monitoring approaches and information about their effects on soil characteristics, including usability, as well as information on possible mobility to other environmental media is presented in this review paper. Finally, a comprehensive overview of remediation strategies which are being developed, including the more traditional ones as well as novel strategies that have been proposed lately by the scientific community, is provided. This includes physicochemical and biological technologies, as well as mixed remediation technologies aimed at enhancing remediation efficiency. Full article

Assisted reproduction technology (ART) has advanced significantly over the past four decades, leading to improved pregnancy outcomes and a reduction in complications, particularly those associated with multiple pregnancies. These improvements largely stem from advances in understanding embryonic physiology, which has enabled better culture conditions. As a result, embryologists can now efficiently culture embryos to the blastocyst stage and successfully cryopreserve them for future use. However, while incubators aim to replicate the maternal environment of the oviduct and uterus, embryos in vitro are cultured in static conditions, unlike the dynamic, constantly changing environment they experience in vivo. Key factors such as pH, temperature, osmolality, and gas concentrations are crucial for establishing optimal embryo development and implantation potential. Moreover, the vitrification procedure for gametes or embryos can introduce oxidative stress, as well as osmotic shock and cryoprotectant toxicity, which may affect embryo viability and increase the risk of birth defects. Since the first successful ART birth in 1978, over 10 million babies have been conceived through these techniques. Although most of these children are healthy, concerns exist about potential birth defects or changes linked to the handling of gametes and embryos. The preimplantation period is marked by significant epigenetic reprogramming, which can be influenced by ART procedures such as ovarian stimulation, in vitro fertilization, embryo culture, and cryopreservation. However, the long-term health implications for offspring remain uncertain. Epigenetic reprogramming during early embryogenesis is essential for proper embryo development and can be changed by ART-related conditions. These concerns have raised questions about the possible connection between ART and a higher risk of birth defects or other changes in children born through these methods. Therefore, we conducted a scoping review following PRISMA-ScR guidelines to map evidence on ART-related risks, including epigenetic and birth defect outcomes. Full article

Every minute, vast amounts of video and image data are uploaded worldwide to the internet and social media platforms, creating a rich visual archive of human experiences—from weddings and family gatherings to significant historical events such as war crimes and humanitarian crises. When properly analyzed, this multimodal data holds immense potential for reconstructing important events and verifying information. However, challenges arise when images and videos lack complete annotations, making manual examination inefficient and time-consuming. To address this, we propose a novel event-based focal visual content text attention (EFVCTA) framework for automated past event retrieval using visual question answering (VQA) techniques. Our approach integrates a Long Short-Term Memory (LSTM) model with convolutional non-linearity and an adaptive attention mechanism to efficiently identify and retrieve relevant visual evidence alongside precise answers. The model is designed with robust weight initialization, regularization, and optimization strategies and is evaluated on the Common Objects in Context (COCO) dataset. The results demonstrate that EFVCTA achieves the highest performance across all metrics (88.7% accuracy, 86.5% F1-score, 84.9% mAP), outperforming state-of-the-art baselines. The EFVCTA framework demonstrates promising results for retrieving information about past events captured in images and videos and can be effectively applied to scenarios such as documenting training programs, workshops, conferences, and social gatherings in academic institutions Full article

Atherosclerosis remains a leading cause of mortality globally, and this is worse in people living with HIV (PLHIV). While the administration of antiretroviral therapy (ART) in this population has significant benefits, it is essential to acknowledge that it also has some undesired effects. This study investigated the impact of ART on carotid intima-media thickness (CIMT) in PLHIV as a marker of early atherosclerosis. A literature search was conducted on the PubMed, Scopus, and EBSCOhost databases from 1 January 1987 to 30 May 2025. The methodological quality of the studies was assessed using the Newcastle–Ottawa scale. Data were analyzed using a meta-analysis web tool and reported as the mean difference (MD) and 95% confidence intervals (CIs). Twenty-seven studies, which included 3250 PLHIV on ART and 1542 who were ART-naive, were relevant. The mean age was 41.26 in ART and 39.91 years. The results showed a higher CIMT in PLHIV on ART compared to the ART-naive group, MD = 0.03 mm, 95% CI (0.02 mm to 0.04 mm), p < 0.0001; I² = 96.9%. Subgroup analysis showed that the inclusion of studies conducted on male participants only, those with a sample size of one hundred, and those with a moderate risk of bias contributed to heterogeneity. The results suggest there is an increased risk of atherosclerosis in PLHIV on ART. Full article

The Seven Masters of the Quanzhen 全真七子 sect served as central figures during the founding phase of Quanzhen Daoism and played key roles in the sect’s early development. Originally positioned as the “Northern Seven Perfected Ones” (Bei Qi Zhen 北七真), they were instrumental in propelling the prosperity and expansion of Quanzhen Daoism. Over time, their images subsequently proliferated across various media—including portrayals in stone inscription, painting, biography, and novel, undergoing transformations through inscriptions, paintings, biographies, and novels—transforming transmission channels from Daoist temples to stage performances and from street corners to modern screens. In the Jin and Yuan 金元 periods, Daoist biographies and inscriptions portrayed the Seven Masters as exemplary figures of Daoist practice. In folk novels and precious scrolls (Baojuan 宝卷) in the Ming 明 and Qing 清 dynasties, they were presented as legendary, divine immortals and distant ancestors available for narrative appropriation. In modern times—particularly due to the popularity of Jin Yong 金庸’s martial art novels—they completed their universalization as Daoist cultural resources blending chivalric ethos and entertainment value. Examining the evolution of the Seven Masters’ imagery, two fundamental implications emerge: First, this transformation was jointly shaped by the power structures, functional needs, and media forms of each era. Second, beneath the fluid representations from sacred patriarchs of the Jin–Yuan period to modern entertainment symbols, there is an enduring thread of Daoist transcendental consciousness. Full article

The rise in social media has improved communication but also amplified the spread of hate speech, creating serious societal risks. Automated detection remains difficult due to subjectivity, linguistic diversity, and implicit language. While prior research focuses on high-resource languages, this study addresses the underexplored multilingual challenges of Arabic and Urdu hate speech through a comprehensive approach. To achieve this objective, this study makes four different key contributions. First, we have created a unique multi-lingual, manually annotated binary and multi-class dataset (UA-HSD-2025) sourced from X, which contains the five most important multi-class categories of hate speech. Secondly, we created detailed annotation guidelines to make a robust and perfect hate speech dataset. Third, we explore two strategies to address the challenges of multilingual data: a joint multilingual and translation-based approach. The translation-based approach involves converting all input text into a single target language before applying a classifier. In contrast, the joint multilingual approach employs a unified model trained to handle multiple languages simultaneously, enabling it to classify text across different languages without translation. Finally, we have employed state-of-the-art 54 different experiments using different machine learning using TF-IDF, deep learning using advanced pre-trained word embeddings such as FastText and Glove, and pre-trained language-based models using advanced contextual embeddings. Based on the analysis of the results, our language-based model (XLM-R) outperformed traditional supervised learning approaches, achieving 0.99 accuracy in binary classification for Arabic, Urdu, and joint-multilingual datasets, and 0.95, 0.94, and 0.94 accuracy in multi-class classification for joint-multilingual, Arabic, and Urdu datasets, respectively. Full article

Physically based realistic direct volume rendering (DVR) is a critical area of research in scientific data visualization. The prevailing realistic DVR methods are primarily rooted in outdated theories of participating media rendering and often lack comprehensive analyses of their applicability to realistic DVR scenarios. As a result, the fidelity of material representation in the rendered output is frequently limited. To address these challenges, we present a novel multi-material radiative transfer model (MM-RTM) designed for realistic DVR, grounded in recent advancements in light transport theories. Additionally, we standardize various transfer function techniques and propose five distinct forms of transfer functions along with proxy volumes. This comprehensive approach enables our DVR framework to accommodate a wide range of complex transfer function techniques, which we illustrate through several visualizations. Furthermore, to enhance sampling efficiency, we develop a new multi-hierarchical volumetric acceleration method that supports multi-level searches and volume traversal. Our volumetric accelerator also facilitates real-time structural updates when applying complex transfer functions in DVR. Our MM-RTM, the unified representation of complex transfer functions, and the acceleration structure for real-time updates are complementary components that collectively establish a comprehensive framework for realistic multi-material DVR. Evaluation from a user study indicates that the rendering results produced by our method demonstrate the most realistic effects among various publicly available state-of-the-art techniques. Full article

Drawing from the insights of community partners, this study explored the roles and benefits of arts-based approaches to foster civic learning, critical media literacy, and community engagement. It also uncovered approaches to promote kindness, prevent violence, and combat online extremism, offering insights into strategies that may enhance community engagement and create a positive impact. We presented our model framework, a detailed case study of our project, and qualitative methods incorporating 15 interviews with our community partners to capture a broad range of perspectives and experiences. Interviewees were community partners who collaborated with our project in organizing events and activities using an arts-based approach to promote kindness, awareness, and violence prevention since the inception of the project. Data were analyzed using thematic data analysis. We categorized the community partners’ responses into four key themes: (1) the inherent benefits of the arts, (2) promoting kindness and preventing violence through artistic expression, (3) teaching civic responsibility through the arts, and (4) practical strategies for collaborating with community partners. The practice implications and lessons learned were discussed. Full article

The present study explores the affordances of virtual reality (VR) technologies to enhance digital and media literacy skills within an interdisciplinary and inclusive K-12 English as a Foreign Language (EFL) learning context. Addressing gaps in research on the design and impact of VR experiences in secondary education, the study investigates VR affordances not only as a learning tool, but also as a medium for knowledge co-creation through learning by doing, with students acting as the agents within digital social contexts. The study was conducted for two years, with 59 participants aged 13–14 years old, following a structured five-phase intervention model with the intent to comply with DigComp 2.2 guidelines for digital citizenship and the Universal Design for Learning (UDL) for inclusive educational practices. The phases involved (a) training on the technological level to leverage digital tools; (b) media and information literacy (MIL) instruction in VR; (c) collaborative VR artifact creation; (d) peer evaluation; and (e) dissemination with peers from other sociocultural contexts for an iterative process of continuous content improvement and social discourse. Mixed methods data collection included pre/post-course surveys, pre/post-tests, observation journals, and student-generated VR artifact evaluations. The findings indicate consistent learning gains across both years, with an average pre–post gain of 18 points (Cohen’s d = −2.25; t = −17.3, p < 0.001). The VR-supported intervention fostered complex skillset building within a VR-supported dynamic learning environment that caters to diverse needs. Students’ reflections informed a framework for designing inclusive media literacy in VR, structured around three main pillars: Narrative Structure, Strategic Design, and Representation Awareness. These themes encapsulate the practical, cognitive, and ethical dimensions of VR design. Sub-themes with examples contribute to understanding the key design elements of VR in promoting participatory engagement, digital and media literacy, critical discourse, and inclusive education. The sub-themes per pillar are signaling and multisensory cues, storyline, and artful thinking; schema formation, multimedia encoding, and optimal cognitive load; and bias-free, respect for emotional impact, and language and symbols. Complementary quantitative findings confirmed the themes of the proposed framework, revealing a positive correlation between the perceived ease of use (PEoU) with digital skills development and a negative correlation between perceived usefulness (PU) and cognitive load. The study concludes with recommendations for pedagogy, curriculum design, and future research to empower learners in shaping sustainable digital futures. Full article

Background/Objectives: This study compares the accuracy of responses from state-of-the-art large language models (LLMs) to patient questions before CT and MRI imaging. We aim to demonstrate the potential of LLMs in improving workflow efficiency, while also highlighting risks such as misinformation. Methods: There were 57 CT-related and 64 MRI-related patient questions displayed to ChatGPT-4o, Claude 3.5 Sonnet, Google Gemini, and Mistral Large 2. Each answer was evaluated by two board-certified radiologists and scored for accuracy/correctness/likelihood to mislead using a 5-point Likert scale. Statistics compared LLM performance across question categories. Results: ChatGPT-4o achieved the highest average scores for CT-related questions and tied with Claude 3.5 Sonnet for MRI-related questions, with higher scores across all models for MRI (ChatGPT-4o: CT [4.52 (± 0.46)], MRI: [4.79 (± 0.37)]; Google Gemini: CT [4.44 (± 0.58)]; MRI [4.68 (± 0.58)]; Claude 3.5 Sonnet: CT [4.40 (± 0.59)]; MRI [4.79 (± 0.37)]; Mistral Large 2: CT [4.25 (± 0.54)]; MRI [4.74 (± 0.47)]). At least one response per LLM was rated as inaccurate, with Google Gemini answering most often potentially misleading (in 5.26% for CT and 2.34% for MRI). Mistral Large 2 was outperformed by ChatGPT-4o for all CT-related questions (p < 0.001) and by ChatGPT-4o (p = 0.003), Google Gemini (p = 0.022), and Claude 3.5 Sonnet (p = 0.004) for all CT Contrast media information questions. Conclusions: Even though all LLMs performed well overall and showed great potential for patient education, each model occasionally displayed potentially misleading information, highlighting the clinical application risk. Full article