Search Results (458)

Multimodal sentiment analysis aims to utilize multisource information such as text, speech and vision to more comprehensively and accurately identify an individual’s emotional state. However, existing methods still face challenges in practical applications, including modality heterogeneity, insufficient expressive power of non-verbal modalities, and low fusion efficiency. To address these issues, this paper proposes a Modality Enhanced Multimodal Integration Model (MEMMI). First, a modality enhancement module is designed to leverage the semantic guidance capability of the text modality, enhancing the feature representation of non-verbal modalities through a multihead attention mechanism and a dynamic routing strategy. Second, a gated fusion mechanism is introduced to selectively inject speech and visual information into the dominant text modality, enabling robust information completion and noise suppression. Finally, a combined attention fusion module is constructed to synchronously fuse information from all three modalities within a unified architecture, hile a multiscale encoder is used to capture feature representations at different semantic levels. Experimental results on three benchmark datasets—CMU-MOSEI, CMU-MOSI, and CH-SIMS—demonstrate the superiority of the proposed model. On CMU-MOSI, it achieves an Acc-7 of 45.91, with binary accuracy/F1 of 82.86/84.60, MAE of 0.734, and Corr of 0.790, outperforming TFN and MulT by a large margin. On CMU-MOSEI, the model reaches an Acc-7 of 54.17, Acc-2/F1 of 83.69/86.02, MAE of 0.526, and Corr of 0.779, surpassing all baselines, including ALMT. On CH-SIMS, it further achieves 41.88, 66.52, and 77.68 in Acc-5/Acc-3/Acc-2, with F1 of 77.85, MAE of 0.450, and Corr of 0.594, establishing new state-of-the-art performance across datasets. These results confirm that MEMMI achieves state-of-the-art performance across multiple metrics. Furthermore, ablation studies validate the effectiveness of each module in enhancing modality representation and fusion efficiency. Full article

Following the significant disruption of traditional teaching by the COVID-19 pandemic, gamified education—an approach integrating technology and cognitive strategies—has gained widespread attention and use among educators and learners. This study explores how game-based learning, supported by situated learning theory and game design elements, can boost learner motivation and knowledge construction. Using 20,293 user comments from the Chinese video platform Bilibili, the study applies sentiment analysis and LDA to uncover users’ sentimental tendencies and cognitive themes. The analysis identifies four core themes: (1) The application of contextual strategies in language learning, (2) Autonomous exploration and active participation in gamified learning, (3) Progressive enhancement of logical thinking in gamified environments, and (4) Teaching innovation in promoting knowledge construction and deepening. Building on these findings, the study further develops a cognitive scaffolding model integrating “contextualization–procedural logic–active construction” to explain the mechanisms of motivation–cognition interaction in gamified learning. Methodologically, this study innovatively combines LDA topic modeling with sentiment analysis, offering a new approach for multidimensional measurement of learner attitudes in gamified education. Theoretically, it extends the application of situated learning theory to digital education, providing systematic support for instructional design and meaning-making. Findings enrich empirical research on gamified learning and offer practical insights for optimizing educational platforms and personalized learning support. Full article

The widespread emergence of multimodal data on social platforms has presented new opportunities for sentiment analysis. However, previous studies have often overlooked the issue of detail loss during modal interaction fusion. They also exhibit limitations in addressing semantic alignment challenges and the sensitivity of modalities to noise. To enhance analytical accuracy, a novel model named MAHFNet is proposed. The proposed architecture is composed of three main components. Firstly, an attention-guided gated interaction alignment module is developed for modeling the semantic interaction between text and image using a gated network and a cross-modal attention mechanism. Next, a contrastive learning mechanism is introduced to encourage the aggregation of semantically aligned image-text pairs. Subsequently, an intra-modality emotion extraction module is designed to extract local emotional features within each modality. This module serves to compensate for detail loss during interaction fusion. The intra-modal local emotion features and cross-modal interaction features are then fed into a hierarchical gated fusion module, where the local features are fused through a cross-gated mechanism to dynamically adjust the contribution of each modality while suppressing modality-specific noise. Then, the fusion results and cross-modal interaction features are further fused using a multi-scale attention gating module to capture hierarchical dependencies between local and global emotional information, thereby enhancing the model’s ability to perceive and integrate emotional cues across multiple semantic levels. Finally, extensive experiments have been conducted on three public multimodal sentiment datasets, with results demonstrating that the proposed model outperforms existing methods across multiple evaluation metrics. Specifically, on the TumEmo dataset, our model achieves improvements of 2.55% in ACC and 2.63% in F1 score compared to the second-best method. On the HFM dataset, these gains reach 0.56% in ACC and 0.9% in F1 score, respectively. On the MVSA-S dataset, these gains reach 0.03% in ACC and 1.26% in F1 score. These findings collectively validate the overall effectiveness of the proposed model. Full article

This article examines the xenophobic orientation of social media reactions, as captured in mainstream South African media, around the Miss South Africa 2024 case of Chidimma Adetshina. It will perform a decolonial interrogation of the South African digital public’s reaction to Adetshina’s participation in and eligibility for the pageant. It will also unpack how xenophobia—defined as the fear or hatred of foreigners—is evident in the backlash that encapsulated Adetshina’s story. The xenophobic utterances that circulated on social media platforms such as X and across different digital media outlets suggest an intriguing intra-black component that is intertwined with the three dimensions of coloniality: power, knowledge and being. The concept of coloniality is understood as the lingering impact of inequalities and power dynamics resulting from the colonial encounter long after the end of administrative and historical colonialism and serves as this article’s theoretical framework. It draws on the work of several decolonial scholars to identify and explore how coloniality presents itself in the Adetshina case. The research objectives are to examine how xenophobic sentiments reflect the coloniality of power, knowledge and, specifically, the coloniality of being. The methodology includes an open, flexible combination of content and textual analysis of online media articles from major news outlets operating within the South African mediasphere. This inquiry found that there is a link between the tension-filled xenophobic reactions to Adetshina’s Miss SA 2024 case and the legacy of exploitation and oppression inherited from South Africa’s still-difficult-to-navigate colonial and apartheid eras. This investigation also found complicated hierarchies between different types of humanity—indicative of the most pervasive dimension, in this case, the coloniality of being. Full article

Artificial intelligence (AI) adoption in strategic sectors such as energy is often framed in optimistic narratives, yet its actual economic contribution remains under-quantified. This study proposes a novel, multi-stage methodology at the intersection of machine learning, statistics, and big data analytics to bridge this gap. First, we construct a media-derived AI Adoption Score using natural language processing (NLP) techniques, including dictionary-based keyword extraction, sentiment analysis, and zero-shot classification, applied to a large corpus of firm-related news and scientific publications. To enhance reliability, we introduce a Misinformation Bias Score (MBS)—developed via zero-shot classification and named entity recognition—to penalise speculative or biased reporting, yielding a credibility-adjusted adoption metric. Using these scores, we classify firms and apply a Fixed Effects Difference-in-Differences (FE DiD) econometric model to estimate the causal effect of AI adoption on turnover. Finally, we scale firm-level results to the macroeconomic level via a Leontief Input–Output model, quantifying direct, indirect, and induced contributions to GDP and employment. Results show that AI adoption in Romania’s energy sector accounts for up to 42.8% of adopter turnover, contributing 3.54% to national GDP in 2023 and yielding a net employment gain of over 65,000 jobs, despite direct labour displacement. By integrating machine learning-based text analytics, statistical causal inference, and big data-driven macroeconomic modelling, this study delivers a replicable framework for measuring credible AI adoption and its economy-wide impacts, offering valuable insights for policymakers and researchers in digital transformation, energy economics, and sustainable development. Full article

This study examines the dynamics of equity market volatility and economic policy uncertainty (EPU) in South Africa during the COVID-19 pandemic. Using daily return data for sectoral indices and the JSE All Share Index (ALSI) from 1 January 2020 to 31 March 2022, the analysis explores both market-wide and sector-specific volatility responses. Univariate GARCH-family models (GARCH (1,1), E-GARCH, and T-GARCH) are employed to capture volatility clustering, persistence, and asymmetry across sectors. The results show that volatility was highly persistent during the pandemic, with sectoral differences in sensitivity to shocks: Consumer Staples and Financials were particularly reactive to recent news, while Health Care and Basic Materials were more stable. Asymmetric models confirm that market sentiment was predominantly driven by negative news, except in the Energy sector, where positive recovery signals played a stronger role. Correlation analysis further indicates that most sectors were moderately correlated with the ALSI, while Energy and Health Care behaved more independently. In contrast, both the ALSI and sector returns exhibited weak and negative correlations with the South African EPU index, suggesting that uncertainty did not translate directly into equity market declines. Overall, the findings highlight the importance of sectoral heterogeneity in volatility dynamics and suggest that during extreme market events, investors can mitigate downside risk by reallocating portfolios toward more resilient sectors. Full article

In recent years, the proliferation of social platforms has drastically transformed how individuals interact, organize, and share information. In this scenario, there has been an unprecedented increase in the scale and complexity of interactions and, at the same time, little to no research about certain fringe social platforms. In this paper, we present a multi-dimensional framework for characterizing nodes and hyperedges in social hypernetworks, with a focus on the understudied alt-right platform Scored.co. Our approach integrates the possibility of studying higher-order interactions, thanks to the hypernetwork representation, and various node features such as user activity, sentiment, and toxicity, with the aim of defining distinct user archetypes and understanding their roles within the network. Utilizing a comprehensive dataset from Scored.co, consisting of more than 4.4 M posts and 36.9 M comments, we analyze the dynamics of these archetypes over time and explore their interactions and influence within the community. We identify eight archetypes, with the largest group comprising over 15,000 users, and observe that 44% of interactions involve at least five participants, highlighting the importance of higher-order modeling. Furthermore, we find significant archetype transitions and stable yet locally dense interaction patterns, with users exposed to roughly 1000 unique peers on average. The framework’s versatility allows for detailed analysis of both individual user behaviors and broader social structures. Our findings highlight the importance of higher-order interactions and node features in understanding social dynamics, and offer new insights into the roles and behaviors that emerge in complex online environments. Full article

In the context of the global climate transition, managerial climate attention is influencing the risk posture of new energy vehicle firms as a key non-economic cognitive factor. This paper investigates the mechanism of managerial climate attention (MCA) on the systemic risk of firms using panel data from 111 listed NEV firms in China from 2013 to 2022. The results show that first, the systemic risk of NEV firms is significantly reduced as managerial climate attention increases. Second, the negative influence of MCA on the systemic risk of NEV firms is more significant among state-owned enterprises, firms in non-first-tier cities and in the machinery, equipment and computer communication sub-sectors. Third, MCA negatively affects the systemic risk of NEV firms by increasing market competition, environmental performance and investor sentiment. The difference-in-differences analysis based on the Paris Agreement shows that the systemic risk of the treatment group enterprises increased significantly after policy implementation, confirming the link between climate-related policies and risk. The management of NEV firms should be concerned about climate change, thus providing practical implications for financial stability and sustainable economic development. Full article

Stock price forecasting remains a central challenge in financial modeling due to the non-stationarity, noise, and high dimensionality of market dynamics, as well as the growing importance of unstructured textual information. In this work, we propose a multimodal prediction framework that combines FinBERT-based financial sentiment extraction with technical and statistical indicators to forecast short-term stock price movement. Contextual sentiment signals are derived from financial news headlines using FinBERT, a domain-specific transformer model fine-tuned on annotated financial text. These signals are aggregated and fused with price- and volatility-based features, forming the input to a gradient-boosted decision tree classifier (XGBoost). To ensure interpretability, we employ SHAP (SHapley Additive exPlanations), which decomposes each prediction into additive feature attributions while satisfying game-theoretic fairness axioms. In addition, we integrate differential privacy into the training pipeline to ensure robustness against membership inference attacks and protect proprietary or client-sensitive data. Empirical evaluations across multiple S&P 500 equities from 2018–2023 demonstrate that our FinBERT-enhanced model consistently outperforms both technical-only and lexicon-based sentiment baselines in terms of AUC, F1-score, and simulated trading profitability. SHAP analysis confirms that FinBERT-derived features rank among the most influential predictors. Our findings highlight the complementary value of domain-specific NLP and privacy-preserving machine learning in financial forecasting, offering a principled, interpretable, and deployable solution for real-world quantitative finance applications. Full article

We propose a two-stage statistical learning framework to investigate how financial news headlines posted over weekends affect stock returns. In the first stage, Natural Language Processing (NLP) techniques are used to extract sentiment features from news headlines, including FinBERT sentiment scores and Impact Probabilities derived from Logistic Regression models (Binomial, Multinomial, and Bayesian). These Impact Probabilities estimate the likelihood that a given headline influences the stock’s opening price on the following trading day. In the second stage, we predict over-weekend log returns using various sets of covariates: sentiment-based features, traditional financial indicators (e.g., trading volumes, past returns), and headline counts. We evaluate multiple statistical learning algorithms—including Linear Regression, Polynomial Regression, Random Forests, and Support Vector Machines—using cross-validation and two performance metrics. Our framework is demonstrated using financial news from MarketWatch and stock data for Apple Inc. (AAPL) from 2014 to 2023. The results show that incorporating sentiment features, particularly Impact Probabilities, improves predictive accuracy. This approach offers a robust way to quantify and model the influence of qualitative financial information on stock performance, especially in contexts where markets are closed but news continues to develop. Full article

Sports teams increasingly use nostalgia-based marketing to spark fan engagement and boost merchandise sales. Yet these efforts can also provoke backlash, especially when they resurrect contested imagery. This article examines how one such campaign—the New York Islanders’ 2015 revival of their controversial fisherman logo—was framed across team broadcasts and interpreted by fans on social media. Drawing on a qualitative textual analysis of television and radio coverage alongside a quantitative content analysis of 563 tweets, the study reveals a divide between institutional messaging and grassroots reaction. While team broadcasts emphasized charity and sentimental appeal, fan discourse was notably more critical, mocking the jersey’s design and recalling past failures. By positioning nostalgia not only as a branding asset but as a reputational risk, the article contributes a novel perspective to debates about commercialization, mediatization, and fan co-production in sports. It also demonstrates the value of mixed methods for analyzing how branding narratives are negotiated in real time. Full article

Understanding how media narratives frame the Sustainable Development Goals (SDGs) is essential for global sustainability governance. This study presents a novel, data-driven analysis of 135,000 news articles mapped to SDGs 1–17 across 100 countries. Using polarity-based sentiment aggregation and principal component analysis (PCA), we reduce high-dimensional SDG sentiment profiles into a two-dimensional space and identify emergent clusters of countries using K-means. To contextualize these clusters, we integrate national-level indicators like Human Development Index (HDI), GDP per capita, CO₂ emissions, and press freedom scores, revealing robust correlations between sentiment structure and developmental attributes. Countries with higher HDI and freer media environments produce more optimistic and diverse SDG narratives, while lower-HDI countries tend toward more polarized or crisis-framed coverage. Our findings offer a typology of SDG discourse that reflects geopolitical, environmental, and informational asymmetries, providing new insights to support international policy coordination and sustainability communication. This work contributes a scalable methodology for monitoring global sustainability sentiment and underscores the importance of narrative equity in achieving Agenda 2030. Full article

The review enumerates the predominant applications of large language models (LLMs) in natural language processing (NLP) tasks, with a particular emphasis on the years 2023 to 2025. A particular emphasis is placed on applications pertaining to information retrieval, named entity recognition, text or document classification, text summarization, machine translation, question-and-answer generation, fake news or hate speech detection, and sentiment analysis of text. Furthermore, metrics such as ROUGE, BERT, METEOR, BART, and BLEU scores are presented to evaluate the capabilities of a given language model. The following example illustrates the calculation of scores for the aforementioned metrics, utilizing sentences generated by ChatGPT 3.5, which is free and publicly available. Full article

China’s rapid urbanization has presented challenges for sustainably revitalizing the historic and cultural heritage within its urban villages. Often, these efforts overlook the crucial roles of community ties and cultural values. This study focuses on 15 representative urban villages in Guangzhou (2019–2024). It tests the core idea that the physical layout of these spaces reflects underlying community structures and cultural values shaped by specific policies. Integrating this understanding into landscape planning can significantly improve revitalization outcomes. We used a mixed-methods approach: (1) Extended fieldwork to understand community networks and cultural practices; (2) Spatial analysis to measure how building density relates to land uses; (3) Sentiment analysis to reveal how people perceive cultural symbols; (4) A coordination model to link population influx with landscape suitability. Key findings reveal different patterns: Villages with strong clan networks maintained high cultural integrity and public acceptance through bodies like ancestral hall councils. Economically driven villages showed a split—open for business but culturally closed, with very low tenant participation. Successful revitalization requires balancing three elements: protecting physical landmarks in their original locations; modernizing cultural events; and reconstructing community narratives. Practically, we propose a planning framework with four approaches tailored to different village types. For instance, decaying villages should prioritize repairing key landmarks that hold community memory. Theoretically, we build a model linking social and spatial change, extending the cultural value concepts of Amos Rapoport to the context of fast-growing cities. This provides a new methodological perspective for managing urban–rural heritage in East Asia. Full article

Sentiment analysis is a cornerstone in many contextual data analyses, from opinion mining to public discussion analysis. Gender bias is one of the well-known issues in sentiment analysis models, which can produce different results for the same text depending on the gender it refers to. This gender bias leads to further bias in other text analyses that use such sentiment analysis models. This study reviews existing solutions to reduce gender bias in sentiment analysis and proposes a new method to address this issue. The proposed method offers more practical flexibility as it focuses on sentiment estimation rather than model training. Furthermore, it provides a quantitative measure to investigate the gender bias in sentiment analysis results. The performance of the proposed method across five sentiment analysis models is presented using texts containing gender-specific words. The proposed method is applied to a set of social media posts related to Morocco’s 2023 earthquake to estimate the gender-unbiased sentiment of the posts and evaluate the gender-unbiasedness of five different sentiment analysis models in this context. The result shows that, although the sentiments estimated with different models are very different, the gender bias in none of the models is drastically large. Full article