Abstract This work presents a modular two layer architecture for recognizing isolated sign language video and converting gloss sequences into fluent natural language text. The proposed pipeline comprises: (1) glosslevel sign recognition using spatiotemporal feature extraction (I3D, EfficientNet,... This work presents a modular two layer architecture for recognizing isolated sign language video and converting gloss sequences into fluent natural language text. The proposed pipeline comprises: (1) glosslevel sign recognition using spatiotemporal feature extraction (I3D, EfficientNet, MobileNet) with lightweight classifiers, and (2) natural language smoothing via large language models (LLMs) using prompt engineering. Unlike prior claims of continuous translation, this paper explicitly focuses on isolated signs as a practical building block for assistive systems where segmentation is either manually provided or handled by external modules. Comprehensive evaluations on Brazilian Sign Language (Libras) datasets demonstrate high accuracy in isolated sign classification (F1 > 0.97 for I3D-RGB + LR). A critical analysis of near perfect AUC scores reveals dataset limitations that are openly discussed. For gloss to text conversion, we evaluate LLM smoothing on both clean and noise injected gloss sequences and report BLEU scores under realistic conditions. The proposed architecture is scalable, modular, and adaptable to other sign languages, advancing accessibility for deaf and hard of hearing communities. Read More Read Less

Abstract The rapid evolution of the Financial Technology (FinTech) sector has intensified market competition and increased the need for enterprises to continuously evaluate and strengthen their competitive advantage. Traditional competitiveness assessment methods often rely on static historical data... The rapid evolution of the Financial Technology (FinTech) sector has intensified market competition and increased the need for enterprises to continuously evaluate and strengthen their competitive advantage. Traditional competitiveness assessment methods often rely on static historical data and subjective managerial judgment, limiting their effectiveness in highly dynamic digital environments. This study proposes a realtime ontology-based machine learning framework designed to evaluate FinTech competitiveness with enhanced accuracy, adaptability, semantic interpretability, and scalability. The proposed framework integrates three interconnected layers: (1) a semantic knowledge representation layer that employs FinTechspecific ontologies to model entities, relationships, KPIs, technologies, and regulatory concepts within the FinTech ecosystem; (2) a real-time data ingestion and preprocessing layer that collects and standardizes heterogeneous data streams from financial statements, digital transactions, customer interactions, APIs, and social media platforms; and (3) a machine learning layer that combines supervised, unsupervised, and streaming learning techniques to classify, predict, and monitor competitive positioning dynamically. The framework is validated using KPI-driven datasets collected from Vietnamese FinTech enterprises between 2019 and 2024. Experimental results demonstrate that ontology integration improves semantic consistency, reduces data noise, enhances feature relevance, and increases model interpretability. The hybrid framework achieved high classification accuracy and effective real-time responsiveness, enabling stakeholders to identify emerging competitive opportunities, operational risks, and market trends. This study presents a scalable and intelligent competitiveness assessment framework that bridges semantic technologies and artificial intelligence to support strategic decision-making in digital financial ecosystems. Read More Read Less

Abstract The exponential growth of scientific literature necessitates advanced computational methods to trace conceptual evolution and semantic relationships. This study investigates N-gram structures and concepttransition hierarchies within 96 research articles from the 2025 issues of the journal Polymers.... The exponential growth of scientific literature necessitates advanced computational methods to trace conceptual evolution and semantic relationships. This study investigates N-gram structures and concepttransition hierarchies within 96 research articles from the 2025 issues of the journal Polymers. Using NLPbased corpus analysis, we evaluated lexical diversity, phrase density, and collocation strength at the unigram, bigram, and trigram levels. Furthermore, we modelled concept expansion trees, semantic pathway networks, and community structures, validating our findings through rigorous bootstrap and permutation testing. The Results shown a progressive increase in lexical diversity and collocation strength from unigrams to trigrams, confirming a systematic shift toward highly specialised, cohesive technical expressions. Core foundational concepts such as "model" and "thermal" exhibit hierarchical expansion into complex trigrams (e.g., "deep learning model"), highlighting the growing integration of computational intelligence into traditional thermokinetics and materials research. Semantic network analysis reveals a scale free architecture dominated by central hub concepts, while community detection identifies four distinct thematic clusters that facilitate targeted interdisciplinary knowledge transfer. Ultimately, the polymer literature corpus exhibits a mature, cumulative, and internally coherent knowledge system. This integrated methodological approach provides a robust, statistically validated framework for mapping the development of scientific terminology, disciplinary maturity, and interdisciplinary convergence in rapidly evolving research domains, offering valuable insights for future Natural Language Processing. Read More Read Less