Search Results (1,531)

Despite significant advances in AI language models, Modern Standard Arabic (MSA) remains a linguistically complex domain in which apparent fluency often masks deeper grammatical instability. This study investigates recurrent grammatical error patterns in ChatGPT-generated Arabic texts, focusing on how these patterns reflect underlying morpho-syntactic challenges and the constraints of probabilistic language generation. Adopting a qualitative, pattern-oriented analytical framework, the study draws on online focus group discussions with secondary-level Arabic teachers, who served as expert linguistic evaluators. Participants collaboratively examined a set of AI-generated texts to identify and interpret systematic grammatical deviations across five key domains: agreement, inflection and case marking, sentence structure, prepositions and transitivity, and cross-linguistic influence. The findings indicate that grammatical errors in AI-generated Arabic are not random but occur as recurring, structured patterns, particularly in contexts involving long-distance dependencies and morphologically complex constructions. These patterns suggest a reliance on surface-level fluency at the expense of deeper grammatical coherence, reflecting limitations in maintaining consistent morpho-syntactic relationships. This study contributes by identifying and characterizing systematic grammatical patterns in AI-generated MSA as interpreted through expert linguistic judgment, offering a qualitative perspective that complements existing quantitative approaches and advances understanding of how large language models engage with morphologically rich languages. Full article

Navigating extensive legislative corpora is often impeded by the linguistic complexity inherent in legal texts. To address this, we present a novel topic representation learning method designed to facilitate the systematic exploration of legislative content. We demonstrate the efficacy of this approach by applying it to the vast corpus of Italian legislation comprising about 74 k laws with more than 300 k articles. While current topic models group documents by latent semantic similarity, they often lack the granularity required for precise navigation. Our approach augments these representations by integrating our topic modeling framework with multi-label profiles. We enrich the representation of individual laws by extracting and ranking the top 10 keywords based on their relevance to the enclosing topic, subsequently aggregating these rankings to construct a comprehensive, alternative description of the broader legal themes. By bridging latent semantic clusters with explicit, LLM-generated labels, this method yields a highly interpretable representation of the corpus, significantly enhancing the profiling and navigability of complex legislative content. We improve over our baseline representation in 74.67% of cases, showing potential for re-use in highly specialized text corpora. Full article

Large-scale datasets such as Objaverse or ShapeNet and repositories such as Sketchfab have been compiled for 3D content. Within the European 3DBigDataSpace project, a consortium of 10 partners assess open licensed 3D models to select and retrieve those models representing cultural heritage objects in Europe to aggregate them into the European Data Space. A key component of this work is the classification and geolocalization of 3D content, with mesh models viewable via different viewers and tested in different scenarios such as museum exhibitions, cultural tourism, or education. This article makes four principal contributions: (1) a current empirical overview of the global distribution and linguistic coverage of large-scale 3D heritage datasets; (2) a comparative evaluation of text-based and image-based methods for geocoding and semantic classification; (3) an analysis of data quality challenges specific to uncurated 3D heritage collections; and (4) a discussion of the implications of user-generated content for definitions of digital cultural heritage. Full article

Fine-tuning pre-trained encoder-based language models for down-stream tasks is typically performed by exploiting the output of the last encoder layer. However, an alternative line of research suggests that leveraging representations from multiple encoder layers may yield richer linguistic information. Previous studies found that different layers convey different linguistic knowledge, suggesting that the last layer might not be optimal for all down-stream tasks. In this paper, we propose a layer-wise attention mechanism using a pivot layer as a new fine-tuning method. The pivot layer is used to compute attention scores of encoder layers, and we define three types of pivot layers. We also examine four attention functions and demonstrate through experiments that the attention function plays an important role in layer-wise attention for fine-tuning. The best-performing combination of our proposed mechanism outperformed the standard fine-tuning method and other recent methods in the General Language Understanding Evaluation (GLUE) benchmark. By visualizing the attention distributions, we found that the last layer is not always preferable for every GLUE benchmark task, and that differences in attention distribution are associated with task performance. Full article

Despite the progressing standardization of the Japanese language, regional dialects persist, particularly among older generations, causing communication gaps, which results in problems especially in healthcare and emergency contexts. This study proposes a text-to-text normalization method to convert eight Japanese dialects into standard Japanese using a fine-tuned mT5-small architecture. We evaluate the impact of learning rate schedulers, training duration, and data preprocessing on model performance. Our results demonstrate that the CharacTER (Character Translation Edit Rate) metric provides a more accurate evaluation than BLEU, which is practically ill-suited for the unsegmented nature of Japanese text. The optimal configuration minimizes character error rates by aligning input data with natural, unspaced Japanese orthography. Furthermore, we observe a statistically significant correlation between the model’s conversion error rate and the physical distance of the source dialect from Tokyo. This finding suggests that the model’s performance effectively serves as a proxy for measuring linguistic distance between dialectal variations and the standard language. Full article

Aero-engine quality assessment under the RAMS framework faces two persistent challenges: the inherent epistemic and linguistic uncertainty in expert evaluation, and the systematic neglect of inter-dimensional coupling. This paper proposes an integrated assessment method that combines Interval Type-2 Fuzzy Sets (IT2FS)-based group decision-making with Partial Least Squares Structural Equation Modeling (PLS-SEM). At the measurement level, IT2FS encodes dual-layered uncertainty through the Footprint of Uncertainty (FOU); multi-expert judgments are aggregated via the fuzzy weighted geometric average operator and defuzzified using the Karnik–Mendel algorithm. At the structural level, a reflective second-order PLS-SEM model built on the RAMS framework enables parametric estimation and significance testing of inter-dimensional coupling. Validation on a 63-engine turbofan dataset confirms that all measurement model criteria are satisfied, the second-order model explains 82.4% of the variance in overall quality (R² = 0.824), and predictive relevance is strong (Q² = 0.567). Comparative experiments against three benchmark methods demonstrate consistent advantages in quality grade discrimination, information richness, sensitivity to technical improvements, and ranking robustness. These properties position the framework as a statistically rigorous, model-based complement to existing condition-monitoring and digital health management systems for complex propulsion systems, supporting quantitative decision-making within digital engineering programmes. Full article

Offensive language detection is crucial for ensuring safe and inclusive digital environments. Identifying harmful content protects users and supports healthier online interactions. Despite advances in transformer-based models, particularly Large Language Models (LLMs), their application to this task remains underexplored for low-resource languages such as Moroccan Arabic, especially compared with high-resource languages. This study evaluates the performance of various open- and closed-source LLMs for offensive language detection in Moroccan Darija. The evaluated models include general-purpose LLMs such as LLaMA, Mistral, and Gemma, as well as Arabic-focused models such as ArabianGPT, Falcon Arabic, and Atlas-Chat. We also experiment with reasoning models such as DeepSeek and GPT-4. Beyond traditional evaluation metrics, we investigate the robustness of these LLMs and examine the impact of adversarial training on their performance. Moreover, we contribute to the field by creating a large, high-quality dataset. Our evaluation revealed that GPT-4o Mini achieved the best overall performance, reaching an F1-score of 88%. However, robustness testing under black-box and white-box adversarial attacks exposed notable vulnerabilities, with attack success rates reaching 30%, thereby highlighting the need for enhancement. Despite the complex morphology and linguistic variability of Moroccan Darija, adversarial training resulted in a notable improvement in both overall model performance and robustness against adversarial attacks, yielding an average increase of 20.89% in resistance to attacks. Furthermore, this approach enabled GPT-4o Mini to achieve an F1-score of 91%, surpassing the current state-of-the-art performance by 6%. These results highlight the importance of incorporating adversarial approaches in low-resource dialectal settings to effectively address linguistic variability and data scarcity. Full article

In this study, we developed a chatbot-based system for detecting early signs of depression and verified its effectiveness through experimental evaluations and user surveys. Emphasizing that it does not rely on medical checklists, the system is designed to automatically extract three linguistic features associated with depression—frequent use of first-person pronouns, pessimistic expressions, and obsessive-compulsive writing styles—from natural user conversations. Multiple models were constructed for these features, and an ensemble layer integrates their outputs for a comprehensive judgment. The implemented system analyzes input sentences obtained through chat, extracts the three categories of features, calculates a final score through an ensemble layer, and visualizes potential signs of depression based on the total score. We performed an evaluation experiment with 20 participants. In the test data evaluation, the system demonstrated over 76% accuracy in each of the three classification categories: first-person usage, pessimistic tendency, and obsessive-compulsive tendency. Full article

This study proposes a methodological framework for extending Malware Information Sharing Platform (MISP) taxonomies in the domain of Dark Web drug forums through the integration of large language models (LLMs) and Human-in-the-Loop (HITL) validation. The research addresses the existing ontological gap between traditional MISP taxonomies, focused on technical or chemical indicators, and the linguistic and morphological complexity of illicit digital markets. By modelling the primary physical form as an ontological predicate with mutually exclusive values (for example, powder, pill–tablet–capsule, liquid, and plant-matter), the proposed approach captures the material dimension of the discourse, enhancing semantic disambiguation and forensic traceability. The Mistral 7B model was used in the morphology-classification stage conducted on a stratified analytical subset of 2904 drug-related Dark Web posts, extracted from a final corpus of 6456 posts after data cleaning and relevance filtering. In the first pass, 76.48% of posts were directly assigned to one of the base morphological categories, while 23.52% were labelled as unclear and subsequently reviewed through the HITL stage. Following HITL refinement and full reclassification, the proportion of posts labelled as unclear decreased from 23.52% to 11.29%, corresponding to a 51.99% relative reduction in ambiguity. Network visualisation with VOSviewer revealed three major discursive axes—recreational–commercial, pharmaceutical–opioid, and transnational–logistical—reflecting the hybrid semantic structure of digital drug markets. The results show that combining LLM-based inference with expert oversight improves the interpretability, reproducibility and ontological robustness of cyberintelligence models, offering a replicable framework for other sensitive domains such as terrorism or child exploitation. Full article

Multilingual learning is key in natural language processing, but is challenged by the transfer–interference trade-off, where positive transfer benefits certain languages, while negative interference affects others. Prior methods, including linguistic-based and embedding-based language clustering, have attempted to address this; yet, they remain constrained by their static design and lack of task-specific feedback. In this study, we propose a novel computational strategy inspired by molecular design that constructs molecules with targeted properties. Languages are modeled as nodes in an undirected graph, with edges representing the transfer strength. This language molecule is optimized via Reinforcement Learning to adjust edge connections and weights to enhance positive transfer and minimize interference, where graph clustering is applied, and clusters are then evaluated on the Named Entity Recognition and POS tagging tasks using 25 languages from the WikiANN dataset and 12 typologically diverse languages from the UDPOS dataset. Compared to linguistic and embedding-based language clustering baselines, our method yields substantial improvements, especially for low-resource languages, with some showing over 35% increase in F1 score, while high-resource languages benefit moderately, confirming reduced transfer–interference trade-off. Our atom–language model offers a novel path for multilingual learning, inspired by molecular principles from physical sciences. Full article

Quality evaluation in production systems represents a significant challenge in the manufacturing industry, particularly in environments where expert judgment plays a key role in managing the inherent uncertainty of the production system. This study proposes a fuzzy multicriteria decision-making control chart, termed Fuzzy Decision-Making Control Chart based on AHP-Extent and Triangular Fuzzy Numbers (FDMCC-AHPE). The method integrates expert knowledge through triangular fuzzy numbers and a Fuzzy Analytic Hierarchy Process supported by Extent Analysis, to define fuzzy decision intervals for quality assessment and subsequently perform a structured analysis to classify the product within a control chart framework. In this framework, expert judgments expressed through linguistic evaluations are systematically translated into triangular fuzzy numbers and processed using FAHP–Extent Analysis, allowing the aggregation of subjective assessments within a structured mathematical decision model. The proposed method was validated in a tannery company, specifically in the retanning process. The industrial case study considers both qualitative criteria, such as surface defects and color uniformity, and quantitative process variables that include bath pH, treatment duration, and processing temperature. The results were compared with an empirical expert-based evaluation and a structured expert assessment supported by a multicriteria decision-making method. The findings demonstrate that the FDMCC-AHPE exhibits greater sensitivity in discriminating between quality states under uncertain evaluation conditions, particularly when samples involve complex evaluation conditions. Full article

The aviation industry operates in a security-sensitive environment where customer feedback may contain not only expressions of satisfaction or dissatisfaction but also threatening or violent language with potential security implications. While conventional sentiment analysis effectively captures customer opinions, it remains insufficient for identifying security-relevant linguistic cues that could signal risks requiring proactive intervention. This study addresses this gap by introducing a security-aware ambient intelligence framework for detecting violent language in airline customer reviews. This framework supports intelligent internet-based monitoring systems and real-time threat detection. We present the first annotated dataset of airline reviews specifically labeled for violent and threatening content, derived from 3629 reviews and balanced through manual resampling to achieve equal representation across positive, neutral, negative, and violent classes. The proposed framework employs VADER-based sentiment analysis for initial polarity estimation, combined with a validated annotation process to identify violent or threat-related content, followed by comprehensive feature engineering combining TF-IDF (2000 features) with text statistics and sentiment scores. We systematically evaluate individual classifiers (Random Forest, Decision Tree, SVM, Naive Bayes) against ensemble methods (Voting, Stacking, Boosting) using accuracy, precision, recall, F1-score, and ROC AUC metrics. Results demonstrate that Stacking achieves the highest raw performance (98.57% accuracy, F1-macro 0.9856), while Naive Bayes offers an optimal balance between effectiveness and computational efficiency (81.79% accuracy, F1-macro 0.8172, training time 0.03 s). This is the first dataset and framework designed for security-aware analysis of airline reviews. The selected Naive Bayes model achieves per-class F1-scores of 0.9978 for neutral, 0.7814 for negative, 0.7482 for violent, and 0.7415 for positive reviews, with a macro-average ROC AUC of 0.7123. The framework is deployed with serialized components enabling real-time prediction, supporting both single-review analysis and batch processing for integration into airline security monitoring systems. This work establishes a foundation for security-aware natural language processing in critical infrastructure contexts, bridging the gap between conventional sentiment analysis and proactive threat detection. Full article

Dialectal variation presents a major challenge for deploying medical language models in real-world healthcare settings, where patient–clinician communication often occurs in regional vernaculars rather than standardized language forms. This challenge is particularly pronounced in the Arabic-speaking world, where clinical interactions frequently take place in diverse dialects that differ substantially from Modern Standard Arabic. Fine-tuning and maintaining separate models for each dialect is computationally inefficient and difficult to scale, motivating more integrated approaches. In this work, we present MENARA, an Arabic medical language model constructed by merging Egyptian Arabic, Moroccan Darija, and medical-domain specialists through model merging. We extend prior feasibility findings through comprehensive evaluation of cross-dialect performance, medical safety, and cross-lingual knowledge retention. Specifically, we introduce a fine-grained dialect composition analysis to quantify lexical purity and structured code-switching behavior, benchmark against state-of-the-art Arabic LLMs, conduct subject-matter-expert assessment of both dialectal fidelity and medical appropriateness. The results show that model merging preserves core medical competence while enabling robust dialectal adaptation, achieving strong cross-dialect fidelity while substantially reducing storage and deployment overhead compared to maintaining separate models. These findings establish model merging as a potentially practical and resource-efficient paradigm for dialect-aware medical NLP in linguistically fragmented healthcare environments. Full article

This study explores how incorporating generative artificial intelligence into the Collaborative Online International Learning (COIL) framework can enhance internationalisation for home and social entrepreneurship education in multilingual settings. A four-week AI-supported COIL programme was conducted with 30 postgraduate students from Russian and Spanish programmes. Students collaborated in intercultural teams to develop bilingual social innovation projects. Data were collected before and after the intervention using validated scales measuring intercultural competence, social entrepreneurship skills, AI literacy and ethics, and linguistic self-efficacy. Repeated-measures ANOVA indicated statistically significant improvements across all domains, with moderate-to-large effect sizes. The most pronounced gains were observed in mixed intercultural groups, which may suggest a potential synergistic effect between authentic intercultural exchanges and AI-mediated language support. Additionally, notable improvements were observed in ethical awareness of AI use and linguistic self-efficacy. Overall, these findings suggest that the AI-COIL model may represent a practical and potentially scalable approach for integrating language learning, intercultural competence, social innovation, and responsible AI use to advance internationalisation in higher education. Full article

Vulnerabilities in conventional agri-food supply chains (CAFSCs) necessitate a shift toward resilient, localized production models. Within the Agri-Food 4.0 landscape, urban Controlled-Environment Hydroponic Agriculture (CEHA) systems address these challenges by shortening supply chains and mitigating climate-induced breakdowns. However, structurally aligning Triple Bottom Line (TBL)-oriented stakeholder needs with complex technical specifications remains a critical challenge in sustainable CEHA system design. To address this challenge, the present study proposes a novel framework integrating the Full Consistency Method (FUCOM) and Quality Function Deployment (QFD) within a Finite Interval Type-2 (FIT2) Gaussian fuzzy environment. This approach systematically translates TBL-oriented priorities into precise engineering specifications, mapping 17 stakeholder needs (SNs) to 30 technical design requirements (TDRs) while capturing linguistic uncertainty and hesitation. The findings reveal a clear strategic focus on environmental and social sustainability. Specifically, high product quality, food safety and traceability, consumer acceptance, and minimization of environmental impacts emerge as the primary drivers of CEHA adoption. The QFD translation identifies scalable IoT infrastructure, sensor maintenance and calibration, and AI-enabled decision support as the most critical TDRs. The framework’s reliability and structural robustness were rigorously validated through comprehensive analyses, including Kendall’s W test to confirm expert consensus, alongside a Leave-One-Out (LOO) approach, weight perturbations, and a structural evaluation of TDR intercorrelations. These findings provide a scientifically grounded roadmap for designing sustainable, intelligent urban agricultural systems. Ultimately, this framework offers actionable managerial implications for agribusiness stakeholders to bridge strategic TBL-oriented goals with practical engineering, significantly enhancing agri-food supply chain resilience. Full article