This study focuses on addressing the growing need for localized language support in socially assistive robots (SARs) due to rising labor costs and the limitations of human labor in developed countries. The research aims to develop a Turkish natural language processing (NLP) module to enhance SARs’ social interaction capabilities and integration into smart living spaces. By leveraging advanced machine learning models, specifically XLM-RoBERTa Large (deepset/xlm-roberta-large-squad2), the study evaluated cross-lingual transfer learning for Turkish question answering, addressing specific linguistic challenges, including agglutinative morphology and vowel harmony. The model was evaluated on the Turkish Question-Answering Dataset (TQuAD 2.0) with 2,520 validation examples, achieving 79.37% F1-score and 56.67% exact match score. The research established a methodological framework connecting adaptive NLP design principles with control systems theory, demonstrating how concepts from adaptive fuzzy control and robust neural adaptive control inform the development of more stable and reliable NLP systems for SAR applications. These outcomes highlight the potential of cross-lingual NLP models for SAR applications in Turkish-speaking environments. The research contributes to the field by: (i) evaluating cross-lingual transfer learning for Turkish SAR applications, (ii) demonstrating the effectiveness of XLM-RoBERTa for low-resource language adaptation, (iii) establishing a framework that connects adaptive NLP design with control systems theory for enhanced robustness, and (iv) identifying real-world SAE applications in healthcare, smart homes, and industrial settings. Future work will focus on integrating this NLP module with speech recognition and synthesis components for complete voice-interactive SAR systems.

Artificial intelligence is increasingly being used as a powerful tool in various industries, including earth sciences. Geological structures play an undeniable role in the formation of mineral potentials. Investigating these structures relies on satellite imagery, together with expert interpretation, which can be a time-consuming process. Artificial intelligence can serve as a valuable tool to expedite this process and enhance the accuracy of mineral potential identification. This article presents a new model based on deep neural networks for identifying mineral potentials. The unique feature of the proposed method is the incorporation of morphological data alongside multispectral data to identify mineral potentials. To evaluate the effectiveness of the proposed method, advanced spaceborne thermal emission and reflection radiometer satellite images from a region in the southeast of Iran were utilized. The results demonstrate an improvement in the accuracy of the proposed method compared to similar approaches.

ynamic hand gesture recognition has become an important research area in human-computer interaction, virtual reality, sign language interpretation, and intelligent surveillance systems. With the increasing demand for natural and contactless communication interfaces, gesture-based systems are gaining significant attention due to their intuitive and user-friendly nature. However, one of the major challenges in dynamic gesture recognition is inter-user variability, where differences in speed, style, and articulation patterns among users reduce the overall robustness and accuracy of recognition systems. Another critical issue is self co-articulation, which occurs when gestures overlap or influence each other during continuous motion, making feature extraction more complex. This study presents a dynamic hand gesture recognition system that addresses inter-user variability in gesticulation patterns. In our proposed system, a new set of features was employed, which divides the gesture into two halves, and feature extraction was performed after the removal of self-co-articulation. The efficiency of the proposed system was validated on a new set of gestures recorded in the LNM Institute of Information Technology Dynamic Hand Gesture Dataset-4, which consists of videos recorded according to different patterns. The performance of the proposed system was calculated with different features combined with individual as well as combinations of classifiers, such as support vector machine, k-nearest neighbor, naïve Bayes, adaptive neuro-fuzzy inference system, and discriminant analysis classifiers. The recognition accuracy of the naïve Bayes classifier was 93.13%, which is the best among all the classifiers. Recognition accuracy improved by about 10% with an increase in the number of features.

Electronic waste (e-waste) is becoming an increasingly significant problem, particularly due to its hazardous and toxic content. The paucity of business management research on technology-based e-waste disposal management exacerbates this issue. The novelty of this study lies in the application of design thinking not to prototype a solution, but to gain a deep understanding of stakeholder experiences and pain points, translating them into structured problem categories that support the development of an ecopreneurial information and communication technology-driven e-waste management system. Understanding the perspectives and experiences of these e-waste stakeholders helps us identify the real issues. In the empathise stage, we conducted interviews with 10 participants, who were chosen to represent the three main stakeholder categories: e-waste disposers, government institutions, and recycling institutions. In the define stage, an affinity diagram was used to analyse the gathered data and formulate the problem. We identified four primary issues with e-waste treatment: inadequate systems, a lack of socialization, low public awareness, and high processing costs. It can be concluded that developing an efficient e-waste disposal management system is a top priority. To realise the possibility of creating an ecopreneur business, further study is suggested to proceed to the ideation stage, where solutions to the four identified problems can be generated.

Blockchain and generative artificial intelligence (GenAI) are two contemporary emerging technologies that have exhibited different adoption trajectories since their inception. Blockchain technology traces its origins to 2008, when it was first conceptualized, whereas GenAI is a more recent development that entered the mainstream with the introduction of ChatGPT by OpenAI. India, as a developing economy, has consistently been at the forefront of technological innovations; however, the adoption patterns for these innovations have been notably different. Using secondary data retrieved from peer-reviewed research and systematic reviews, along with industry and market intelligence reports, this research revealed that blockchain, as a technology, adopts a bottom-up approach driven by financial inclusion imperatives and is inherently decentralized by design. GenAI, on the other hand, adopts a top-down approach, fueled by enterprise-driven adoption and rapid scaling across various sectors. Our findings suggest that the difference in their diffusion approaches is attributed to the persistent regulatory uncertainty and infrastructure constraints faced by blockchain, whereas GenAI has benefited from clearer policy support and lower entry barriers. This paper provides a frameworkbased, side-by-side comparison of two high-impact technologies in a single national context, linking micro-level adoption mechanisms to macro-level diffusion outcomes. These nuances could have significant implications for policymaking and recalibrating India’s position in the global landscape.